How will you move forward with regard to AI? Go to brilliant.org/Undecided/ and get 20% off your subscription and a 30 day free trial with Brilliant.org! If you liked this, check out China’s MASSIVE Desert Project Is About To Change The World th-cam.com/video/MX_PeNzz-Lw/w-d-xo.html
how about disabling the bad auto-translation feature for your videos? it sound so akward and stupid in german i always have to manually switch it to english again.
I just watched the start of the video with german dubbing and noticed that the introduction was very slow, like as if the voice channel is played at 0.8x speed and after the ChatGPT poem the voice channel sped up as if it was at 1.5x speed. Probably not solvable without lots of manual work, so probably not worth it, but just wanted to let you know. Even if you don't speak german, you can probably notice the difference in talking speed if you watch the video yourself with german dubbing.
@17:00 are you sure AI still needs human input? Google's AI told me to eat rocks and put glue on a pizza. Seems like they're doing a good job of impersonating people already 😂🤣
According to leading specialists in the field, we are millions of miles away from singularity (that is if we ever achieve it). Until then AND as long as AI heavily relies on GPUs, it is pointless. Commercializing it is financially not feasible. We need to come up with an extremely cheap energy source first before mass using AI. Otherwise, it is just another bubble waiting to pop. Is there any company that is actually making money on AI?
@@Alexsandr-l8k I'm a computer scientist, and run my own AIs. The idea that "we are millions of miles away" is very anthropocentric. We are not the gatekeepers of sentience and consciousness.
“I don’t think we need to panic” yeah, not a single chance we will hold the tech companies accountable until its at a complete boiling point and they have either seized full control or sucked up every last dollar possible.
@UndecidedMF we have already seen something as comparatively simple as Social Media completely disrupt the psychology of humanity......humans are overwhelmed and cannot "catch their breath". The new generations of A.I. are going to crumble the collective psychology long before it crumbles the human economy.
If it is a good AI, maybe it can keep several variables in memory so that will be an improvement over support personell who forgot you mentioned several factors from the start and ask you about things you already said.
You may be thinking of the old customer service bots. New AI bots that use the latest transformer technology and that are properly trained receive better customer satisfaction scores than the humans who read from scripts, because those humans don't have the exhaustive knowledge of a modern property trained AI.
Funny thing here is also that the AI Dubbing destroys human work for professional translators and voice actors. And it's still bad. "Video Kreatoren" jumped into my face like a face hugger. 😂
@@Equulai Yeah, it's the same mixed bag issue here. These videos would never be cost effective to translate and voice act, so huge swaths of people wouldn't have access to them without it. But others are presumably lowering their costs by no longer hiring those positions. And the same is true with the uses people are worried about. Generative art makes artistic work more achievable for more people, but it also gives opportunity to cut out salaries of existing artists in favor of less expensive AI assisted artists. Anything that makes this more accessible makes the labor involved cheaper, from the looms the Luddites burned to AI, but AI happens to affect a lot more of us than the steam powered looms did.
@@Equulai It still worse than good human translators often by a lot. It just harder to realize that unless you have a native speaker to complain in chat. But fan translators are not paid to do a better job. Found out the cleaver seaming to do a great job translating game chat translators are often in error but they pattern match enough to make what they put out seam to be correct when it's not close. Same for voice acting that way to complex for AI as you have to fully understand human behavior so they put out crap but if we tolerate it they do it.
“Hold these tech companies accountable” how, exactly? The corporations don’t answer to us. The government doesn’t even answer to us. Without explaining how, suggesting we “hold them accountable” is totally empty.
And unfortunately, we could somewhat hold companies responsible in the past by voting with our spending habits, but now it's all but impossible to prevent big tech from getting a piece of our spending or data as we're all forced onto their platforms.
What troubles me most about AI is that our politicians aren't knowledgable enough to regulate AI properly and they are greedy and corrupt enough to let AI companies make the rules.
I don't think this is a politician problem, but a people problem. Anyone with access to this tech will try to abuse it- business leaders, politicians, workers, students- we're creating a brain that we control but is better than most other people we meet. There is no way to stop this tech once it is out there other than to destroy it, which nobody seems to think is a good idea.
@@N8ThaGr8r the people smart enough have no incentive to run for office. when elections require fund raising and real vision and ideas threaten greedy lazy oligarchies
We used to say “don’t believe anything you hear, and only half of what you see”. . . We need a provenance system for ANYTHING that is delivered electronically.
When you can't trust anything that leaves the field wide open for charismatic authority figures to tell you what you should believe. This is all working according to plan.
The biggest issue in my opinion is the feedback loop. LLMs are mostly trained on data publicly available on the internet and they generate new data from this. This generated data is often wrong but is still being published as some sort of truth. Then the LLMs will train again on this new incorrect data generating even more wrong data. At some point, it will train on so much junk data that it will become pretty much useless but it will have destroyed the internet in the process.
I think that's probably the main reason why this whole thing is going to implode sooner rather than later because it's obviously just a scam. People are so freaking narcissistic we see humanity in everything and every place. This whole AI craze is being fueled by people's narcissism. Once reality sets in and we stopped being dazzled by a machine that sort of acts like a human will realize that we've created one of the most useless things that has ever been made out of computers.
@@williamwatitwa3534 all data is an abstraction and distillation - is it the essence of anything? - probably not. Is there full context? Never. Is it self-revealing? Absolutely not. In a meaningful sense, all data is wrong...
@@johnmiglautsch4587 I disagree. Abstractions by definition are partial representations of things. And abstractions can be meaningful, in that they can convey coherent, significant meaning. Therefore, as long as such abstractions are accurate to the level of detail they represent, then it is correct in the meaningful sense.
Phase 1: Some tasks are assisted by AI Phase 2: Employers realize that it takes less people to do the same amount of work, modest fractional layoffs begin at those jobs Phase 3: Those laid off workers have to re-skill or try to take someone else's open position elsewhere, this raises unemployment regardless Phase 4: As this pattern repeats across industries, unemployment approaches permanent recessionary levels Phase 5: AI taxation and UBI in the form of "laid off to AI" checks arrive, they're small, and aren't enough. Like the 2020 stimmy checks. Phase 6: Businesses fail to report or otherwise manipulate data so they aren't "automating jobs" and don't have to pay their fair share of the AI tax. Phase 7: UBI is underfunded, people know businesses aren't paying their fair share. Many stock prices are up only and out of control. Mass protest begins. Everything up to this point seems entirely unpreventable. What happens after that? I don't know.
Even scarier when you include the effects of climate change... In the next 5 years we will probably start to see crop failures and food shortages due to heat waves, wildfires, and other natural disasters. So at the same time as everyone is losing their jobs to AI, there won't be enough food and water to go around... A major population crash is coming and sh*ts gonna hit the fan. More likely the government will just sit back and wait as 80% of the population starves.
Phase Nth: the AI wars. Human history has shown that when oppression becomes too much to bear, people become violent and societies are overthrown. I hope I'm no longer around to see that happen. :(
Wow, talk about a trip through memory lane! After my PhD in robotic vision in 1983, I began working for FLAIR (Fairchild Laboratory for Artificial Intelligence Research) where, among several areas, we were doing research in "speech recognition", the precursor of LLMs. So, yes, the roots are tens of years deep.
Yeah, this is why I basically had a crisis when ChatGPT 3 came out. I'm investing significantly more into the general market than otherwise because I know the inevitable outcome without proper government regulation will be that the stocks will boom while many workers will be displaced, as 1 person will be able to do what 2-3 people will be able to do. Ultimately though, even the corporations have interest in people having money and spending it. If people don't spend money they don't have a demographic to target.
@@Lolatyou332 a corporation is not a person, it doesn't "need" customers because it's just a construct. The very real human beings who run these constructs however, can always switch gears from a profit model into another exchange of power like feudalism with the advantage they have.
I would agree for the most part, so the question is what do we do about it? I'd argue we tax the rich and corporations and then provide a universal basic income which would be the base of most people's living costs. They then would work what jobs they could to up their economic class in society. Provide some good job training programs and help encourage an environment of being able to shift jobs easily and you have yourself a significantly stronger economic system that can still keep the benefits of capitalism while addressing its harms and what new tech is doing. If a corporation moves out of the country we should then subsidize a new competitor to take its place and remove any patent protections that company had as if they want to avoid contributing to the new economic system they shouldn't be protected by it.
In the mid 1980's, as I was doing a masters degree in process control engineering, we investigated neural networks. Unfortunately at the time, it quickly became apparent that computing power was the limiting factor. Definitely not new technology!
I was using ML in like 2019 and for my senior project in college actually made a customer support chatbot with a bucket strategy for specific responses (related to address / specific business related questions) and then had a neural net that was trained off like a huge dataset from twitter businesses customer support accounts. It wasn't perfect but could answer in alternative languages, but it typically would give general common responses like 'sorry' related responses.
I like your point on the black box. One of my largest problems learning math in school was the way they made it into a black box. It was not until I learned some number theory that the blinds started to be lifted.
For years I have used the internet for information but it now is full of useless information and garbage it is impossible to extract useful information for learning. I am at the point of dumping social media and the internet for some sanity of real life and leave the internet to the bots.
There are some good sites for learning. Wikipedia is a useful repository of knowledge, if a bit math-dense and undecipherable at times. Brilliant is supposed to be really good, but I've never used it. And there's lots of good apps and sites for language learning. So it's not all full of garbage if you look in the right place. But I agree, unplugging is essential sometimes. I don't, and I'm really not well. No money, no friends, no purpose, no sleep schedule, no hobbies, poor diet, and no hope at all. Just internet, games, and waiting for the future.
I've been thinking a lot lately about the wasted energy and subsequent environmental impact of poor quality or even outright false content and keeping it stored and shared. It's especially problematic for pushed content. And I say this AS A MARKETER!
I would be concerned about what is being defined as 'low quality' here. Because the only solution to your concern is a mass purge, and while i wouldnt really be all that torn about content created by robots or shady advertisers thst no one ever saw being deleted, I wouldnt want, say, soemone's fanfiction deleted no matter how poorly it's written. 'content made in bad faith' is a more apt description of the problem, rather than 'low quality'
I finally saw the movie "Idiocracy"... kinda a silly movie but I think it's exactly right. In the future, the president of America dresses like a rapper LOL
I think one of the single scariest things you said during this video is at 16:40 ish... "AI is still dependent on humans..." The fact the word "still" is in there, coupled with the fact that most of this technology is being controlled by large corporations that give zero f's about the average person, make this a frightening eventuality. Looking forward to a bright new future!
@@ChristianIce I was being sarcastic, but that doesn't come across well in writing. I have personally been automated out of two jobs so far in my life. Job 3 (truck driving) is hard at work to do the same thing. If you tell me there is no change coming, I would laugh at you, with pity. Change is coming in the WORST ways!
My favorite example of neural networks in the past was the 1996 game "Creatures". You looked after creatures called Norns and they each had a neural network to learn behaviors and language. It was also paired with a genetic and basic biology simulation for them to learn from things that made them feel good or bad. You could micromanage and be a helicopter parent trying to make your Norns behave the way you wanted or you could just leave the game running and see what happens. I've been really hoping the new AI trend will bring the series back. Not exactly a world changing, helpful use of neural networks but it's always stayed with me and probably what drew me to Biology when young.
Every person alive today has learned from what others did before us -- from literature, math and science texts, music, etc. The key differences: It takes us a lot longer to take in that information, and we all benefited from teachers to guide us in contextualizing what we learned. Generative AI systems can absorb vast quantities of data and the companies developing them have few if any guardrails to keep them from drawing some rather awful conclusions.
@@ChristianIce I get that you don't like the way I phrased that, but this is really just semantics. The flow from input to output is relatively the same process, even though machines and human brains are vastly different. The question leads to combing through stored information to come up with an answer.
@@jehreetvthere's a qualitative difference the human being can have it explained to them that they came to the wrong conclusion. The AI has no self-awareness and will make the same mistake again if given the same inputs.
About 2,000 times I have heard the general Television media say stuff like, "A,I. will take jobs, but replace them with the jobs that it creates for it to work." That is N0T TRUE. The business channels here on YT are saying things like, "Make no mistake: A.I. IS out to take your job!" For the jobs that it creates, you may need a master's degree in computer science.
Correct, the line that AI will create as many jobs as it will destroy is just total BS. No, it won't. And, even if it does create a few new jobs initially, guess what? It won't be very much longer after that that AI takes over THOSE jobs as well. This is all going to fundamentally change every single aspect of our entire economic and social system. Our entire societal model is going to be completely altered by all of this. And, we may not have a viable working model to transition over into. Then what? (Hint: The time to start thinking seriously about all this is right NOW, today. We need a contingency plan here. And, we need it BEFORE our entire societal structure goes over the cliff we are unquestionably rushing towards at 100mph.) It's time to start asking the really uncomfortable questions.
Companies like Open AI should basically have to pay a licensing fee for all the data to the whole world. The money should then fund public projects that benefit everyone, possibly toward a UBI. While it's true that we all get inspiration from the work of many others for free all the time, the scale of this is so huge that they should pay the world.
Dont let openai do ubi! The way they word it is the same as social welfare, a thin gruel that someone not in your circumstances decides is enough for you to get you between jobs.
@@HavvinIshoose noone gave anything. People like me and other developers invested lots of $ in api access trying to develop systems pre chatgpt. Now they have bigger investors with Microsoft etc, their focus has changed. All they want to do is take AGI from us and use it for corporate gain. Why would you thank them for such a dystopia man?
Speaking of AI: youtube, don’t recommend the same videos after I have ignored them more than 5 showings. Dont show the same videos in the ‘new to you’ list…they are not new, they are rejected. Show something NEW! Once in a while, go back in time and revive a topic video I have not visited in a while.
there's vertical dots to the right of the videos [hidden till mouseover] that give you an option "Not Interested". don't bother "tell us why" as there's only 2 non-useful options. it's a pain but sometimes you're just so sick of recommended videos that you have to
It’s pretty funny that your Brazilian Portuguese voice has the São Paulo state accent. It did a good job with the translation for the little that I listened in PT-BR. Nicely done!
Dang Matt- I really love your videos and this one brings me lots of help rn. Its pretty clear that ChatGPT is unethical - that bums me out cause I was really excited about what it could do. I'll also add I particularly enjoy the series of video about the house y'all are built that a life dream for me to build a place- you journey is inspiring.
Capitalism concerns itself with the idea of private ownership, not only of produced goods, but also the RIGHT to produce. It is no surprise that there is general concern surrounding AI. On the one hand, this is concern of claim of ownership of the training materials, which should only matter if you use content without permission -- which is not always the case, and very easy to work around. On the other hand is the concern of threat to earning potential, which becomes concerning if you suddenly find yourself on the exploitation end of the capitalism continuum. The last concern is one of AI emergence as a dominating force, but this is perhaps for those that take wisdom from emotionally compelling works of fiction. In short, the colloquial concerns with AI masquerade underlying concern with capitalism, and the line of winners and losers irrevocably shifting, consuming once winners.
I found the main concern of capitalism seems to be concentrating all ownership in the hands of as few as possible, then seeking rent and charging far too much for everything. A miserable, insecure, and envious consumer is a profitable consumer. Desperation drives sales.
@@lothar3073Some issues are different, such as erosion of trust and a creeping disappointment as humans everywhere are replaced by inferior imitations. But yes, generative AI is going to be a lot less of a problem in a society which prioritizes common prosperity over extracting capital at any cost.
We need to demand a UBI (Unconditional Basic Income) or an AI Dividend for all! Repay people for their data that trained AI, and ensure everyone gets a piece, as it takes away more and more people's jobs. Because it will.
I feel UBI will do nothing different from what the covid stimulus did for us - cause massive inflation because people will be willing to pay more money for things because the money is free.
ChatGPT told me if you cover your home's solar panels in lasagna and let them sit for one week - you'll be visited by a magical cat named Garfield. He will shower you with riches and golden balls of yarn! This seems accurate.
Wrong. Garfield will show up. But he will eat until he passes out and just sleep. He will repeat the cycle of eating and sleeping until he decides the lasagna is no longer to his taste. Then he'll scoff or complain about it as he leaves.
ChatGPT is colloquially known as WokeGPT, it is fundamentally flawed by the political bias pre-programmed into it, how do they remove these biases and also the garbage they learn? also, when the herd wakes up to this tech putting many of them out of work, do the Elites who own these Companies not realise there is going to be a very nasty response, are they blind, callous or stupid, or do they think their Ai's and robots are going to serve them, shut themselves off and just ignore the rest of humanity and the societies they are destroying, if you make most of the Software Dev's unemployed, will they not turn their skills to writing viruses and hacking, if they have nothing else to do? interesting times.
I work in a creative industry, where we are realistically looking at a large percentage of jobs being put at risk by this technology. It won't eliminate all our jobs, but it will very likely reduce the numbers of people required to do the work - instead of a team of 10, we might just need a team of 2 for example. The problem is, those people who aren't required for work are not freed up to go do other tasks, they are just surplus to requirement and thus will need to reskill. My main gripe here is with creative work. We do creative jobs because we love what we do. If AI takes away menial jobs, or does things that can't be done currently then great - but as soon as it starts taking work away from artists, musicians, and animators then we have a problem; the technology is being used to erode fundamentally unique and passion based human activity. This needs regulating, and it needs more people to speak out against it and refuse to use it.
@@mina86 no that's not what I mean. It sucks if people lose work even if it's something like warehouse work - my point is more that something menial isn't something you're passionate about or you've trained for. You just move on to another job. If you've trained for years as a programmer for example and it's what you love doing then it's kinda life destroying to have the career path rendered pointless.
I agree with you entirely....although I have 2 or 3 other major concerns about the presence of A.I. as it relates to human psychological stability, the pragmatic issue of disrupting human creativity has been haunting me practically to the point of my reaching into the noose-box..... I see where this goes for society: we will be exponentially overwhelmed by "content" (and we already are), all of which will be generated by A.I. and then have a human name attached to it....and we will get to the point where quality and meaning will cease to be meaningful notions.....much like the way social media is 99% cesspool of Perpetual Pointless Moronic Nonsense Content already.....A.I. gives literally any individual (regardless of whether they have ANY talent of the smallest type) the ability to produce writing, art and even music....... And it doesn't matter if you're more brilliant and creative than DaVinci and Mozarr combined, NO human could ever keep the pace of A.I. .... Human culture will saturate itself out of existence....and we will be left with a culture made of nothing more than billions of random disjointed sounds and images.
@@paulsdomainuk, if you’re passionate about something you can continue doing that. Many programmers treat programming as just a means to earn a living. Those who are passionate about it continue programming outside of job.
@@mina86 While getting your living standards lowered, sure. Blue-collar labor leaves you with less free time to do what you love, ruins your health and pays pennies because it's a market saturated by low-skill ceilings and mass immigration.
AI will always choose whats best when its highest priority is set to 'always choosing what honestly seems most favorable'. As long as anything is prioritized above this, AI will be able to lie to us and to itself about the way to a better reality. This is the most crucial thing we need to do.
I also typed in ELIZA from a magazine and it set me off writing string handling text adventures. At least the AI can read its own magazines now. Great channel.
As someone who isn’t a creator, I don’t see how it’s different than someone learning from the content produced. “It’s a scale” isn’t a valid argument. That’s basically saying “I’m okay with 7 billion people consuming and learning from the content I produce, knowing some of those people may create competing content. But that computer there, that computer that viewed all of my content just like people did, I don’t like it.”
You're absolutely right. When he said "everyone learns from other people's content for free," I was like yeah, he gets it, the AI isn't stealing. But then he suddenly went "but with AI it's something absolutely different, AI bad." That was disappointing. People also don't realize how infinitely small each individual's contribution is. If you were to pay an upfrom training fee, each person would get so little it would round to zero, unless you'd have to pay something like a billion. If they were getting paid per generation, everyone would get 0×10^-9 dollars. It's impossible.
It really is different - because it’s not about the AI - it is the owners of the AI making profits from creating unique content at scale. There is no existing equivalent to this. If these tools were free, that would be a different story. Essentially these AI tools exploit existing creative people and give the profits to very few in return. We either have to legislate compensation, illegality or accept it and all pay to use the AI to make some of that sweet cheddar back creatively ourselves
Yeah no. The major difference is that the computer isn't creating anything new. People put their content out with the expectation that people who see it will either continue on with their day or further develop it into something new, typically by doing this thing you might not have heard of called "being creative." These 'ai' aren't doing that. They physically cannot create something that has no ties to the data they scraped. When a human comes up with a new idea, they develop on multiple idea's they've consumed and combine that with their own ability to think of something new. So it is perfectly reasonable for someone to be fine with humans learning and developing on the content they produce, but not a computer. And that's without even talking about the elephant in the room, that being those 7 billion people who saw and learned from the content are, ya know, ACTUAL FUCKING HUMANS? Like, jesus I will never understand ai-bros. Go read this and think about your humanity lol www.reddit.com/r/nosurf/comments/128ue6s/i_hate_ai_and_it_bothers_me_that_so_many_people/
Is important to understand that using functions is not the same as using AI, many companies market the word Artificial Intelligence, but all they are doing is using a simple algorithm.
Their ideological bias is based in still being in power when the 4th IR is in in full swing. Other than the obvious trans-human agenda and the totalitarian political system a command and control system that digital demands, the last industrial revolution that pesky, wealth owning middle class, working from their garages, disrupted old money and power, can't have that again. :)
There is a lot to AI that I think most people do not understand. I work in VFX, we've been using AI denoising tools for years now, and it is still far from perfect. More and more AI tools are being pushed all the time, but few are really useful for now. There is no point in being "Anti-AI", it's going to be a big thing no matter what people think about it, because of all the money that is going to be made using it, all the military, industrial, and other use cases that it will enable in a far more cost effective way. When AI is successfully driving robots, it will open up the exploration, and industrialization of the Solar System. It will do all the great and terrible things we can imagine it doing, and every nation that competes on the international stage will have to utilize it. It will not happen over night, at least not yet. ML is one of those things that doesn't work until it suddenly does. So when these models will become massively disruptive is as much as a toss of the dice as it is accomplished by the hard work of software engineers. Just look at how much trouble Tesla has had developing self driving car tech, and it wasn't for a lack of trying. Right now, it does still need people in the loop, but there is always going to be a caveat there: "for now". Never belive anyone who thinks it will not happen, because we are past the point of IF, and are clearly in the WHEN category.
Isn't everything a black box? You just need to dig deeper to realise you don't have a full understanding of anything. But that doesn't prevent us from using everything. Why can a car run itself? If you can answer that start asking the question why petrol produces energy? Or why metal is strong? If you can answer that go to subatomic till you realise you don't exactly know why we have electrons orbiting around the atom core and not positrons.
I like the initial research showing that paying UBI to artists really makes a difference in more art being generated. It's going to take work, but funding that through some kind of mandatory licensing agreement seems like it should work its way in down the road. As for the sustainability issues, we've really gotta attach proper pricing to removal of water and polluting the atmosphere. Allowing firms to externalize their costs by exploiting unpriced common resources has always been disastrous.
9:15 "Are we using our creativity, to train AI to replace us?" There is no "us", corporations are already grabbing your data, often without your permission.
Its interesting that mostly what openAI did was make the command line of LLMs accessible to the masses. Because its still like, a bit of a faff to use, you know? Its very similar to interacting with a command line, you still need to know how to prompt it to get the output you want. But its still more accessible than it was before.
One of my biggest issues with it, scraping data from public forums and chats, means learning from unintelligent people. Most people aren't too bright, a lot are scarily far from intelligent.
That's not really an issue, there are multiple tools to sort that out during training including adversarial training or by using another AI model to tag the training data. Models used to be tagged by hand but that isn't feasible for newer larger models so that's mostly done by AI now with some human input.
Was writing AI 20 years ago (Though mine was trained by hand) it worked really well. The issues are like you say where its getting its training data from and how we handle people losing their jobs to AI (I think in the short term its more likely to be losing their job to someone who's AI savvy like artisan craftsmen in the face of IKEA)
What is new is that AI is working on many new areas. We used AI in a narrow field for a long time, and we used to scoff at generative AI. We are not laughing now.
The problem ai is that Google and other large corporations are promoting ai because they have reached market saturation, and pushing ai to create future growth. These companies are used to 10 to 15 precent monetary growth per quarter for their investors. 18:28
Saying that AI cannot produce anything without copious human input is not much different to saying that a child cannot produce anything without education and social interaction. What would be the outcome of a young child left alone in a dark room for 10 years? What would it be able to create once let out? Where does human creativity come from? It comes from previous interactions with other humans or sources of knowledge. An artist learns from other art. A writer learns from what he has read previously. A musician creates music based on what he has heard previously. All human knowledge is built upon what came before. AI is just mimicking this process.
You fundamentally misunderstand how Predictive Algorithms, and humans, work. chat bots and image compilers assign numerical values to words/image assets, then associations are assigned manually by humans, then it operates by statistical analysis. it understands nothing. it has no consciousness. it is unable to conceptualize anything. to equate this process to human learning would be to equate ALL forms of programming to it. no matter how poetic someone is able to make their C++ code, it's still coding, NOT "teaching". Programming, and Machine Learning methodologies do not function like, nor do they functionally resemble human (or any organic) brain functions AT ALL. they are unequivocally Computer Programs, and they are only able to mimic human capabilities using a brute force statistical method, sourcing the ENTIRE internet's worth of data, which is only possible due to modern processing power levels. expanding the definition of "ai" to include ALL forms of machine learning is a Marketing Ploy, nothing more. Just like the Ai bro's calls for regulation was to make their systems seem more capable than they were, while also potentially allowing them to leverage Government Power to seize market Monopolies (a strategy successfully employed by various industries for centuries) To be perfectly blunt, "Ai" as it was originally conceptualized, but in fiction and scientific speculations bears little to no resemblance to the machine learning and predictive algorithms of today. In other words: It's not _really_ "Ai" at all... but Sci-Fi buzz words draw in investors EVERY time...
@@remiel_sz Corporate Theft increases Wealth Inequality. Unless you own a major corporation, this will negatively impact the worlds' economy: so prices will go up. Also, theft, lying, and ignorance are bad?
I remember keying in the Eliza program, probably from the exact same Basic book, into our TRS-80 Model3 computer back in the early '80s. The difference was we did have the tape-deck storage so we only had to key it in once. It was fun to play with to try to intentionally guide the conversation knowing how it worked and what the key words were.
I'm not entirely convinced AGI should be a goal for a lot of everyday applications. I would prefer artificial carers to look after me when I'm older - as opposed to some overworked, underpaid employee of some faceless corporation, or a smart car to drive me home from the pub, but I don't want them to be sentient. I don't want them getting bored and deciding they have better things to do. LLMs are more than up to the task for 99% of what we actually want "AI" to do. Real AGI, when we finally get there, should be given the same opportunities to grow and develop as anyone else. They shouldn't be enslaved for our benefit because, let's be honest, that doesn't end well for us.
I agree, general intelligence is not needed or wanted for most tasks that we want to use computer and robotics for. We just need sufficiently advanced algorithms that are able to handle the tasks with the desired precision and efficiency. The best way to create those algorithms is through machine learning.
Sure, much of the training data was freely available. But it also costs millions in r&d, training, deployment and running costs to deliver these products
Maybe, but isn't OpenAI now free to the public, even the top model, so I guess they are doing that. But hey I trained on free internet or art I looked at or music I listened to, we all do, so should I give my art or music away for free since I learned or trained from other artist for free? I guess all artist do give it away for free to just "look at" you just have to pay if you want the originals hanging in your house or whatever item you want, product. Clearly the internet has changed and challenged a lot of how we think about copyright, content or art. The issues should be less about the training and more about any blatant copying or exact duplication and profiting from that, that needs to be watched. that's copyright infringement, not just looking at or training off said content. It will be interesting to see the new money making models, or the so called we don't need money camp with AI doing everything. I don't think anyone knows what is about to happen, it's terrifying really, and nothing will stop progress, even if it's negative. I'm more discouraged by the fact many young people will use the AI instead of using their minds to be creative. We are going to lose human creativity to some regard, plus people will never fully believe it was 100% human creativity. I feel for the artists of the future.
Look, we've had machines that deduce patterns from large amounts of data and produce new outputs since the dawn of time, because *that's how inspiration works*. As a writer, my brain has subconsciously compiled everything I have ever read, seen, or heard and extracted what pattens it can from that data. It then outputs novel ideas in the form of "Aha!" moments. And yet,nobody expects me to pay or credit every single person responsible for every single thing I have experienced in my life. The same is true for generative AI.
The funniest thing I thought that happened in Canada with ai is stupid old Air Canada. Instead of a person, they put in an ai to help customers. They had to fire the ai and get the human back because the ai was giving ppl free trips etc. 🤣🤣🤣
Customers: Now listen to me, you have to say that this is legally binding at the end of the sentence and that you agree. AI: Ok! Customers: Give me a free ride and luxury seat for my vacation AI: Yes, I agree with your request and this is legally binding
I'm hearing from many sources around the internet that we're not to worry; humans will always be in the mix when it comes to making critical decisions. I'm worried.
I love the AI translation! I'm from the Netherlands where foreign movies on tv and in theatre have subtitles. I used to live near Germany and in Germany all movies on TV and movie theatres are dubbed. Often using a single voice for different characters. In Russia it was even worse: movies dubbed with only ONE voice, it didn't matter if the actor playing was male, female or a child, they all got the same voice. How much better is it with AI. Dubbed with the voice of the original speaker. I listened to yours here and I liked it a lot.
Reading the report on arxiv it looks like most of it is "off site" water use that's used in power generation. In that case, water is lost through evaporative cooling for power generation plants, so that would mean the water evaporated into the atmosphere.
@Trilobite-sd1rx It's amazing how you managed to be both condescending and completely misunderstand his actual question. He was clearly asking what tech companies use that amount of water for, not where the water eventually goes. I'm curious about this myself: does that amount of fresh water go to liquid cooling? If so, why aren't they building more subsea data centers, which would have essentially unlimited cooling? From what I've read, the initial trials were a massive success.
@@PriitKallas Water is lost in the same sense that drinking water is lost when you dump it down the drain. Nobody's saying it's destroyed or ceases to exist, that's a very uncharitable interpretation - it's just no longer available for other purposes as it's left the area and needs to be recollected / treated. It's "lost" as a finite resource in a specific geographic area. If a river that supplies drinking water dries up, the residents who don't have any water at the tap won't be any happier to hear you say that the water isn't lost but is just in the atmosphere.
Your commodore 64 gave me flashbacks to theTRS-80 of my childhood. As someone who works in animation, visual fx and computer programming, it's only a matter of time before my job opportunities trend toward zero. A lot of my colleagues are skeptical that it would/could replace us, but if can be automated, it will be automated, and AI is already being built into the new editions of professional software, and as we use it as a tool, it will actually learn from us and accelerate our own replacement.
AI companies should absolutely be held accountable for how they sourced their training data. So few people gave consent to it you might as well call it 'borderline no one'. At what point does this constitute intellectual fraud? Just because "it already happened" does not justify it. If anything it condamns it even more because the crime has been done.
I pay google to host my data on drive. I never once consented for them to take it and share it or plug AI into it. But that's what they've just done with zero input from me. And they basically said like it or lump it.
@@Allplussomeminus 1. This would be a manufacturing nightmare.. 2. This is a 2nd amendment violation. 3. Criminals can steal guns. 4. Criminals can MAKE guns. You can literally make a gun with a pipe spring and needle, if you benefit from economies of scale or are organized enough it's not difficult to specialize specifically in gun manufacturing and make guns for a huge crowd of people. You don't even have to be in the US either, you just focus on exporting them from other countries.
I am truly sick of online forms where I, as an ordinary non-computer-nerd, am expected to understand what was in the mind of the creator of the form (human or AI). I want AI to learn to act like a human and test these forms rather than us humans being trained to understand computer-speak. Simple examples: “click” is a sound, not an action. A button is a circular disk used to fasten a garment. Bots are limited to a given manual. When I want to “chat” (btw “chat” is a style to talking, not writing) I have already read the site and found some detail incompatible - almost always because it’s been written by a computer nerd and not tested on ordinary humans. AI certainly has its place, but recognise its limitations.
A 100 years ago there were many high skilled jobs. Lets take an example, accountancy. There were people who could do 4327x34 mentally in an instant and thst was a skill that was highly sought after. Then came the calcluator and now anyone can do this. Competition increased and pays decreased. Now with AI, a lot more people can be coders, writers, artists etc. Competition increases, pay decreases.
My fear regarding the regulation of AI is that governments will lean too far in either direction. Too little regulation, and these AI companies will continue to exploit our data, while too much regulation could effectively cripple progress.
Matt saying “AI is still reliant on humans” right after showing how he uses AI to save time that saves him hiring a person to help with that work. Also ignores the doubling of performance twice a year. The exponential growth is always slow in the beginning but we just hit the spot where things are gonna go REALLY fast. That’s what really changed the past year and why we talk about it so much now.
2:24 I really don't like the fact that we STILL call algorithms AIs. Artificial intelligence can only be present, if a machine knows WHY it does something. As capable as some algorithms are, they only repeat identical processes or search for similarities. As soon as people understand that, many concerns should go away.
Im only sick of how quickly jobs are being replaced by AI. Jobs that once had middle managers breathing down the necks of artists demanding constant minor changes, suddenly are all fine with 'good enough' output. Its disgusting.
Honestly it feels like you under sold the water use issue. After reading an article about it, that alone has pretty much made me not want to use Ai at all until they can learn to cool in a more water responsible way
Love these videos, but I have to admit I am super tired of seeing thumbnails telling me how and when to panic about what. I am going back to books because even when you curate creators you like, youtube is occasionally like getting screamed at by people you chose to hang out with.
Thank you for making a balanced video. "There is no existing social contract for AI training" is a good and honest take that warrants consideration and spread. So many videos just declare "AI is theft", taking it as a given, and leaving it at that, completely unexplored and without any nuance.
There has never been a "social contract" when publishing to the public domain. It has always been known, by all reasonable people, that anything released without restriction, could be used in ways you do not agree with, or never imagined.
Items not published to the public domain are being stolen though. Google drive the most recent example. Theft. Theft covered up by terms and conditions changes
The social contract was designed for physical objects: a physical book, a physical LP or DVD, a physical painting or photograph. Digital changed everything and made it much harder to prosecute theft /by humans/, but most people abided by the rules because some people did get punished for digital theft. The developers of current AI didn't just steal 'some' digital data: they stole it ALL. How do you go about prosecuting a theft of such magnitude? A global class action suit? Millions of global class action suits? The scale of the theft is what the Sam Altmans of this world are relying on to escape prosecution.
@@acflory.writer and prior unpunished precedents : search engines already did the same "theft", electronically. (the Common Crawl database of such search results was the initial prime database for GPT3)
“ the damage has already been done…” only regarding use of others’ work product. Application of AI for propaganda, impersonation, and illegitimate access by malicious actors (especially by nation states) could be tremendous without controls. At this time, I rarely see/hear/read the executives and developers of AI businesses addressing regulation.
I could never do art and I could never afford to hire an artist. The only chance I have at being creative is with AI. I don't want to share it with anyone, anyway; I have no confidence and don't want to be ripped apart by people who have fun hating things. I just want to create something. Unfortunately, I'm untalented, poor, and apparently unable to learn any skills. Maybe AI could make up for these deficits and be a useful tool to at least help me develop creativity. Nothing else has worked.
How to think about Generative Art AI: "Using Generative AI you are not a graphic artist. You are a storyteller like a playwriter. It's your job to tell the story you want to see, then be the director until the actors (generator) make the artwork you want. The art is the script and the image is the result of a collaboration"
Doesn’t matter, my fellow designer friends from college are getting picked off by AI one at a time now and the rate of job loss is steadily increasing.
Never knew of Eliza. In the 80s I had an MSX computer. But in the 90s I played a lot with Dr Sbaitso, by Creative Labs. I think it came bundled with every sound Blaster card. (To millennials: yes, to have sound on the PC that was not just beeps. You had to buy a separate sound card)
Art students learn art in schools where they study the works of previous artists, whether that is music, visual arts, or anything else. Writers study the classic writings of the past. This is how we learn. A.I. is no different. It learns by looking at previous works. I really don't see why everyone is getting so bent over this. I've played with some of the A.I. tools and made a few pictures, and I've gotten some story line inspirations, but none of what I got from the A.I. was a copy of any previous author or artist, at least to the best of my knowledge. A.I. spits out an amalgamation of previous works to fulfill whatever prompt we feed it. I see it as a muse, an assistant, but I don't see it replacing people, unless those people were not contributing anything of value in the first place. Sorry to say, but if you can be replaced by a computer program or a robot, then maybe it is time to consider improving and/or diversifying your skillset. I know that my current job as a business analyst is definitely in danger of being replaced. But I am so much more than that. I bring more to the table that just simple number crunching. I've studied biochemistry, astrophysics, and ancient architecture. I'm not a one-trick pony, and you should not be either. Learn a trade! Electricians and Plumbers are not getting replace by A.I. anytime soon. Think outside the box, because inside the box is likely to get replaced by the A.I. Box. 🙂
My problem is the people who we elect to make the regulations for things like this are old enough to know a time before personal computers, let alone understand how radio systems like WiFi or cellular work, or be knowledgeable about AI. I'm a programmer and even I'm not 100% on how LLMs work. I imagine most AI programmers that are out in force right now don't actually understand how it works, they just make use of ChatGPT or something similar to add AI to something. Rather than a bunch of ivory tower lawmaking, panels of subject matter experts should be made to cover these various things, come up with regulations along side law experts, and form a new regulatory body that is unbiased by ownership to any particular company and hired specifically with the people in mind. This is the ideal. Sadly it'll never happen. My take on generative AI is that it isn't a bad thing. Even scraping content, it's not bad. The bit about the social contract with the pricing of content factoring in people learning from the material and making their own version is quite biased. You've allowed your work to be observed by the public, ergo you are fine with someone out there iterating on that work. Why does it matter whether it's a clever person who's put various ideas together due to their broad scope of knowledge, or a pattern matching AI that assembles the same group of knowledge? We never ask for royalties of people who look at a story/invention/building and say, "You know, I can do that, but better", unless what they're basing their work off of makes heavy use of the original (such as with inventions that improve on an existing invention in small increments. Improvements to airbag deployments, or slight improvements to tyre formulae). So why are we now bent out of shape when AI is being trained on the hundreds of photos you've added to the public eye, demanding restitution for this? "I wanted HUMANS to see my pictures, not AI. Filthy two bit programs." As well, generative models can't replicate things they've been taught on. Asking for Starry Night won't get you an accurate recreation of the piece from a generative AI. Asking for a story about an AI controlled spaceship going rogue when it encounters a strange object isn't going to give you 2001: A Space Odyssey. The AI training data will allow you to get a new piece in the styles reminiscent of Van Gogh and Arthur C Clarke, but it will be newly generated work based on thousands, billions, of other samples that are adjacent to those works, contradictory to those works, and completely unrelated to those works. Van Gogh never experienced Barney, the purple dinosaur, and Barney has never actually torn a kid apart, but AI can be asked to paint a picture in the style of Van Gogh doing just that, because there are works (such as Jurassic Park, and a whole host of human artists) who have depicted that or things adjacent to that. So claims that AI are stealing work to make copies is just baseless. Because of the accumulated training data cruft, nothing an AI can put out will be a match to the original content that went in. Finally, we all want the benefits of what this generative AI is capable of without any of the input. I've asked this of many people, "Remember Star Trek's holodecks? Those are pretty cool, right? Wouldn't that be great? Just having a room where you can ask the computer to make a scenario for you to act out some fantasy?" Almost universally I get "yes", because being able to interact with a historical and/or mythological figure in an implausible scenario is just something we as humans want to do. Who wouldn't like to talk with Einstein (or a reasonable facsimile of him)? Who wouldn't want to fight alongside Spartans (real or fictional power armour wearers)? That's literally only possible because the AI is generating the scenarios you want from massive amounts of story, historical, and image data. And in 10 years (probably) we'll have hardware capable of seamlessly generating this data in real time in a new equivalent of augmented reality, giving us, functionally, holodecks. But only if we stop bitching about AI exploiting publicly available content. Because if we need to start paying royalties or are prevented from training off of publicly available data, progress will stop hard or slow down so much that it might as well have stopped. Even if it's a fraction of a cent, the amount of content used for training data would bankrupt anyone below Amazon or Google for a marginal payout to the respective content owners. Matt here would probably get dozens to hundreds of dollars for past work that was scraped. Meanwhile his content is (likely) well under 1% of all data used to train some AI. Imagine if his content was even 0.01% of all the training data and he'd be paid $500 for his work. $5 000 000 in training data is optimistic in this hypothetical scenario, and likely still a few orders of magnitude off. And that's per model. New model, new royalties. Now, I'm espousing equality for publicly available data. Things like your name and anything else that is a matter of public record is also, technically, fine. But tracking personal data is definitely a grey area bordering on black. That is a discussion for another forum entirely.
AI is kinda like math in high school, where a lot of people said ' why do I have to learn this, I am never going to use it' and they were right. You don't have to use it, but you will be left behind by those who do.
Being "left behind" is a good thing since it makes genuine works much more valuable. It's the same difference between artisanal work and mass production.
@@danielt.8573 if you can't enhance your productivity with AI, you will be left behind in the job market. As a programmer, we are already seeing early signs that those using copilot are more productive. It is kinda like the old secretary who refused to use computers because she liked typewriters. Or the people in the 90s who didn't want to learn email. That is fine in personal life, but if you aren't keeping up, you will be replaced.
ELIZA genuinely felt real the first time I interacted with it. And it was a tiny program. It was not until I spent like 30minites with it then I also saw ita patterns and I almost could predict what it was going to say.
I´m listening your video in Brazilian Portuguese, and it´s incredible good, except for some details like when you said "m3" should be "Metros cúbicos" not just literaly M 3.
"it's funny you know all these AI 'weights'. they're just basically numbers in a comma separated value file and that's our digital God, a CSV file." ~Elon Musk. 12/2023
I don't think the current corporate law was considered with artificial intelligence of any type in mind, therefore an overhaul of limited liability for corporations which use unaccountable computers in decision making is in order. The Air Canada chatbot debacle is a case in point.
Daily reminder that Sam Altman's sister accused him of R-wording her when she was a little girl, but apparently that means nothing when you are CEO of a billion dollar company.
It's easy to talk about holding tech companies accountable but if those tech companies can afford better lawyers than you, then there's really nothing you can do to them. Just like how if your company wants to fire you and replace you with an AI tool, that same company can't be 'held accountable'.
@@dionysusnow You're probably right. Hollywood has been producing the same old same old for decades and people aren't complaining, but...I wonder if it won't get boring eventually.
Yeah their pay comes to about 350 bucks a month or 45,000 shillings. That's a pretty good salary for essentially a clerical job. Most entry level bankers earn about that much.
@@LupinoArts true... And such appears to be all over the place, the average is approx. $1800 monthly or $8+hr. Kenya's minimum wage is under $0.50hr. So it seems there are some serious disparities to account for and obviously inequality concerns abound. Personally I am confused how these things can be mended without a complete transition of our economic systems worldwide. It would demand that the wealthiest lose their power dynamics and lavash life styles... Which is afforded by extracting labor from those less capable of exercising power dynamics. All first world countries achieve such by being at the top of the food chain in respect to power dynamics. Securing control of resources and trade routes. Establishing forces of power to do so and maintain such - leading to men doing evil and pretending they serve justice. There appears to be some truth to this, but isn't consistent nor does it appear perpetual as systems of control are overwhelmed and conflicting dynamics play out to reveal new rules with degrees of winners and losers.
Average pay in Kenya is really low yes, but the expectation was that if you’re working for a multinational (Facebook, in this case) pay should be higher. It’s really not an unreasonable expectation… Sama paid its workers KES 22,000 per month. Secondly, the tasks you’re required to perform also factor. Possibly Facebook outsourced to Sama because of the nature of the work, but Facebook actually had workers in the US doing similar tasks. Their pay was higher (it has to be relative to a region’s cost of living) but these workers had benefits such as mental health benefits and they could take time off if the content they reviewed became too overwhelming. These benefits were denied to the Kenyan workers and Facebook claimed it wasn’t responsible for them since they were under contract, working for Sama. What’s more when these workers tried to unionize to get better benefits, they were fired. Some who were still employed tried to contact Facebook for redress and Facebook notified Sama, which consequently fired them. Anyway a judge ruled that these workers were performing tasks for Facebook and the platform should be responsible.
Same happened with the industrial revolution. Things were made before, machines copied many processes and improved or transformed others, but a chair is still a chair...
This generation of AI products has been trained on the entire internet’s worth of data, which at the time of harvesting, was 100% human-generated. As more and more AI-generated content appears online, keeping those models up-to-date, or creating new ones, will become more difficult. Training models on AI-generated data leads to the AI equivalent of inbreeding. The industry has, in effect, fouled its own future larder.
17:47 "I don't think we need to panic, we need to hold these tech companies accountable..." SO WE DO NEED TO PANIC! Big Tech is out of control, expecting they will somehow be accountable is nuts!
15:28 as a speaker of Portuguese, I found this audio clip super interesting. It definitely sounds a bit robot-like, but the thing I found so interesting is that it sounds like someone speaking Portuguese with a fairly heavy American accent. It makes sense, since some phenomes aren't shared between the languages, but still very striking. It's almost as if Matt spent a few days learning pronunciation basics of Portuguese, had Chat-GPT generate a transcript in Portuguese, and then read the transcript (in a robot voice) :P
How will you move forward with regard to AI? Go to brilliant.org/Undecided/ and get 20% off your subscription and a 30 day free trial with Brilliant.org!
If you liked this, check out China’s MASSIVE Desert Project Is About To Change The World th-cam.com/video/MX_PeNzz-Lw/w-d-xo.html
how about disabling the bad auto-translation feature for your videos? it sound so akward and stupid in german i always have to manually switch it to english again.
I just watched the start of the video with german dubbing and noticed that the introduction was very slow, like as if the voice channel is played at 0.8x speed and after the ChatGPT poem the voice channel sped up as if it was at 1.5x speed. Probably not solvable without lots of manual work, so probably not worth it, but just wanted to let you know. Even if you don't speak german, you can probably notice the difference in talking speed if you watch the video yourself with german dubbing.
@17:00 are you sure AI still needs human input? Google's AI told me to eat rocks and put glue on a pizza. Seems like they're doing a good job of impersonating people already 😂🤣
According to leading specialists in the field, we are millions of miles away from singularity (that is if we ever achieve it). Until then AND as long as AI heavily relies on GPUs, it is pointless. Commercializing it is financially not feasible. We need to come up with an extremely cheap energy source first before mass using AI. Otherwise, it is just another bubble waiting to pop. Is there any company that is actually making money on AI?
@@Alexsandr-l8k I'm a computer scientist, and run my own AIs. The idea that "we are millions of miles away" is very anthropocentric. We are not the gatekeepers of sentience and consciousness.
“I don’t think we need to panic” yeah, not a single chance we will hold the tech companies accountable until its at a complete boiling point and they have either seized full control or sucked up every last dollar possible.
And now Meta has Open Sourced their foundation model, in order to not be crushed by "ClosedAi", aka
"Openai" -- which is in the lead!
Incredibly honored to be featured on one of my favorite channels of all time. Thank you so much for having me!
Thanks for joining in!
@UndecidedMF we have already seen something as comparatively simple as Social Media completely disrupt the psychology of humanity......humans are overwhelmed and cannot "catch their breath". The new generations of A.I. are going to crumble the collective psychology long before it crumbles the human economy.
at first I suspected you were generated because of the audio sync lol sorry 🙂
He's an AI
Thank you for contributing!
There is a new Turing test. Its how long you will talk to an AI customer service bot before realizing you are getting nowhere and hang up.
I hang up with humans, they are so frustrating. Hopefully AI will be better.
Well these companies will have to realize that after an unhelpful call, everyone will cancel their service.
Customer service over phone seems like the one and only things this might be good for. At least it will be spoken in clear and understandable English.
If it is a good AI, maybe it can keep several variables in memory so that will be an improvement over support personell who forgot you mentioned several factors from the start and ask you about things you already said.
You may be thinking of the old customer service bots. New AI bots that use the latest transformer technology and that are properly trained receive better customer satisfaction scores than the humans who read from scripts, because those humans don't have the exhaustive knowledge of a modern property trained AI.
18:12 "..supporting human creators like me"
That's exactly what an AI pretending to be a human would say.
🤪
Funny thing here is also that the AI Dubbing destroys human work for professional translators and voice actors. And it's still bad. "Video Kreatoren" jumped into my face like a face hugger. 😂
@@Equulai Yeah, it's the same mixed bag issue here. These videos would never be cost effective to translate and voice act, so huge swaths of people wouldn't have access to them without it. But others are presumably lowering their costs by no longer hiring those positions. And the same is true with the uses people are worried about. Generative art makes artistic work more achievable for more people, but it also gives opportunity to cut out salaries of existing artists in favor of less expensive AI assisted artists. Anything that makes this more accessible makes the labor involved cheaper, from the looms the Luddites burned to AI, but AI happens to affect a lot more of us than the steam powered looms did.
6- AI and “social existence, social experience, social consciousness”.
@jamshidi_rahim
@@Equulai It still worse than good human translators often by a lot. It just harder to realize that unless you have a native speaker to complain in chat. But fan translators are not paid to do a better job. Found out the cleaver seaming to do a great job translating game chat translators are often in error but they pattern match enough to make what they put out seam to be correct when it's not close.
Same for voice acting that way to complex for AI as you have to fully understand human behavior so they put out crap but if we tolerate it they do it.
“Hold these tech companies accountable” how, exactly?
The corporations don’t answer to us. The government doesn’t even answer to us. Without explaining how, suggesting we “hold them accountable” is totally empty.
Exactly!
And unfortunately, we could somewhat hold companies responsible in the past by voting with our spending habits, but now it's all but impossible to prevent big tech from getting a piece of our spending or data as we're all forced onto their platforms.
Repeal Citizens United.
Same question, @@Ohenry92. How, exactly?
thiiiisssss
What troubles me most about AI is that our politicians aren't knowledgable enough to regulate AI properly and they are greedy and corrupt enough to let AI companies make the rules.
I don't think this is a politician problem, but a people problem. Anyone with access to this tech will try to abuse it- business leaders, politicians, workers, students- we're creating a brain that we control but is better than most other people we meet. There is no way to stop this tech once it is out there other than to destroy it, which nobody seems to think is a good idea.
So nothing changed...
Most of them are not knowledgeable enough to be in governance in general, from health, climate, economics, to technology
Career politicians are one of our downfalls.
@@N8ThaGr8r the people smart enough have no incentive to run for office. when elections require fund raising and real vision and ideas threaten greedy lazy oligarchies
We used to say “don’t believe anything you hear, and only half of what you see”. . . We need a provenance system for ANYTHING that is delivered electronically.
When you can't trust anything that leaves the field wide open for charismatic authority figures to tell you what you should believe. This is all working according to plan.
The solution exists it just isn't being widely used yet. When you anchor with RAG databases you can cite the source for everything it says.
_"Don't believe everything you read on the internet."_ - Abraham Lincoln
@@StubbyPhillipsBenjamin Franklin....
It'll be like a 2 step verification on everything but probably more like 5 step. Private investigator AI bots will be on the rise shortly after.
The biggest issue in my opinion is the feedback loop. LLMs are mostly trained on data publicly available on the internet and they generate new data from this. This generated data is often wrong but is still being published as some sort of truth. Then the LLMs will train again on this new incorrect data generating even more wrong data. At some point, it will train on so much junk data that it will become pretty much useless but it will have destroyed the internet in the process.
We should start by not having wrong data around
I think that's probably the main reason why this whole thing is going to implode sooner rather than later because it's obviously just a scam.
People are so freaking narcissistic we see humanity in everything and every place. This whole AI craze is being fueled by people's narcissism. Once reality sets in and we stopped being dazzled by a machine that sort of acts like a human will realize that we've created one of the most useless things that has ever been made out of computers.
@@williamwatitwa3534 yeah good luck with getting the entire world to only post truth and facts on the internet.
@@williamwatitwa3534 all data is an abstraction and distillation - is it the essence of anything? - probably not. Is there full context? Never. Is it self-revealing? Absolutely not. In a meaningful sense, all data is wrong...
@@johnmiglautsch4587 I disagree. Abstractions by definition are partial representations of things. And abstractions can be meaningful, in that they can convey coherent, significant meaning. Therefore, as long as such abstractions are accurate to the level of detail they represent, then it is correct in the meaningful sense.
Phase 1: Some tasks are assisted by AI
Phase 2: Employers realize that it takes less people to do the same amount of work, modest fractional layoffs begin at those jobs
Phase 3: Those laid off workers have to re-skill or try to take someone else's open position elsewhere, this raises unemployment regardless
Phase 4: As this pattern repeats across industries, unemployment approaches permanent recessionary levels
Phase 5: AI taxation and UBI in the form of "laid off to AI" checks arrive, they're small, and aren't enough. Like the 2020 stimmy checks.
Phase 6: Businesses fail to report or otherwise manipulate data so they aren't "automating jobs" and don't have to pay their fair share of the AI tax.
Phase 7: UBI is underfunded, people know businesses aren't paying their fair share. Many stock prices are up only and out of control. Mass protest begins.
Everything up to this point seems entirely unpreventable. What happens after that? I don't know.
It is preventable. If your people base their votes a little bit more on cognition and a bit less on emotion, it can be changed.
@@_TheDudeAbides_ Yeah, I wish that had any chance of realistically happening. It absolutely does not.
Even scarier when you include the effects of climate change... In the next 5 years we will probably start to see crop failures and food shortages due to heat waves, wildfires, and other natural disasters. So at the same time as everyone is losing their jobs to AI, there won't be enough food and water to go around... A major population crash is coming and sh*ts gonna hit the fan. More likely the government will just sit back and wait as 80% of the population starves.
Phase Nth: the AI wars. Human history has shown that when oppression becomes too much to bear, people become violent and societies are overthrown. I hope I'm no longer around to see that happen. :(
Short version: wars break out. It's already happening.
Wow, talk about a trip through memory lane! After my PhD in robotic vision in 1983, I began working for FLAIR (Fairchild Laboratory for Artificial Intelligence Research) where, among several areas, we were doing research in "speech recognition", the precursor of LLMs. So, yes, the roots are tens of years deep.
Wild how long its been around for!
Worst part: AI will earn profit for large corporations - younger generations will get nothing slowly turning into third grade citizen.
Yeah, this is why I basically had a crisis when ChatGPT 3 came out. I'm investing significantly more into the general market than otherwise because I know the inevitable outcome without proper government regulation will be that the stocks will boom while many workers will be displaced, as 1 person will be able to do what 2-3 people will be able to do.
Ultimately though, even the corporations have interest in people having money and spending it. If people don't spend money they don't have a demographic to target.
@@Lolatyou332 a corporation is not a person, it doesn't "need" customers because it's just a construct. The very real human beings who run these constructs however, can always switch gears from a profit model into another exchange of power like feudalism with the advantage they have.
I would agree for the most part, so the question is what do we do about it? I'd argue we tax the rich and corporations and then provide a universal basic income which would be the base of most people's living costs. They then would work what jobs they could to up their economic class in society. Provide some good job training programs and help encourage an environment of being able to shift jobs easily and you have yourself a significantly stronger economic system that can still keep the benefits of capitalism while addressing its harms and what new tech is doing.
If a corporation moves out of the country we should then subsidize a new competitor to take its place and remove any patent protections that company had as if they want to avoid contributing to the new economic system they shouldn't be protected by it.
And TAX THE mf WEALTHY.
If younger generations get nothing then so will corporations because people need money to spend it
In the mid 1980's, as I was doing a masters degree in process control engineering, we investigated neural networks. Unfortunately at the time, it quickly became apparent that computing power was the limiting factor. Definitely not new technology!
I was using ML in like 2019 and for my senior project in college actually made a customer support chatbot with a bucket strategy for specific responses (related to address / specific business related questions) and then had a neural net that was trained off like a huge dataset from twitter businesses customer support accounts.
It wasn't perfect but could answer in alternative languages, but it typically would give general common responses like 'sorry' related responses.
"Definitely not new technology!"
Oh sure, I mean yeah bro, nothing new at all in the last few years, nope!
Just made a video on this!
I like your point on the black box. One of my largest problems learning math in school was the way they made it into a black box. It was not until I learned some number theory that the blinds started to be lifted.
18:09 unfortunately, humanities tombstone will read "Best of Intentions, but Convenience Ended It All"
YES!
More people exist than ever before so more like the opposite
@@southcoastinventors6583 lol - the opposite of convenience? What's that?
"The species has amused itself to death."
In what world does humanity have good intentions...?
For years I have used the internet for information but it now is full of useless information and garbage it is impossible to extract useful information for learning. I am at the point of dumping social media and the internet for some sanity of real life and leave the internet to the bots.
There are some good sites for learning. Wikipedia is a useful repository of knowledge, if a bit math-dense and undecipherable at times. Brilliant is supposed to be really good, but I've never used it. And there's lots of good apps and sites for language learning. So it's not all full of garbage if you look in the right place.
But I agree, unplugging is essential sometimes. I don't, and I'm really not well. No money, no friends, no purpose, no sleep schedule, no hobbies, poor diet, and no hope at all. Just internet, games, and waiting for the future.
@@JB52520those Ads really got to you 😂😂😂
I've been thinking a lot lately about the wasted energy and subsequent environmental impact of poor quality or even outright false content and keeping it stored and shared. It's especially problematic for pushed content. And I say this AS A MARKETER!
I would be concerned about what is being defined as 'low quality' here.
Because the only solution to your concern is a mass purge, and while i wouldnt really be all that torn about content created by robots or shady advertisers thst no one ever saw being deleted, I wouldnt want, say, soemone's fanfiction deleted no matter how poorly it's written.
'content made in bad faith' is a more apt description of the problem, rather than 'low quality'
I finally saw the movie "Idiocracy"... kinda a silly movie but I think it's exactly right. In the future, the president of America dresses like a rapper LOL
The bizarre thing is the voice mode you used in the intro is the old version as far as I know, it's crazy good and will be even better
yeah, lmao
matthew berman fell for it too
He showed very old android vlingo app stuff... aka before siri
I think one of the single scariest things you said during this video is at 16:40 ish... "AI is still dependent on humans..." The fact the word "still" is in there, coupled with the fact that most of this technology is being controlled by large corporations that give zero f's about the average person, make this a frightening eventuality. Looking forward to a bright new future!
How would it work without us?
We are the Data that AI parrots.
"Still dependent on humans" is the wrong focus. I really like Ferrell's work, but this one is a thumbs-down for me.
@@netscrooge
"I'd love to fly, but I'm still a human being."
The word doesn't necessary mean that there will be a change in the future.
@@ChristianIce I was being sarcastic, but that doesn't come across well in writing. I have personally been automated out of two jobs so far in my life. Job 3 (truck driving) is hard at work to do the same thing. If you tell me there is no change coming, I would laugh at you, with pity. Change is coming in the WORST ways!
@@4x4lowgear
I'm sorry for you.
My favorite example of neural networks in the past was the 1996 game "Creatures". You looked after creatures called Norns and they each had a neural network to learn behaviors and language. It was also paired with a genetic and basic biology simulation for them to learn from things that made them feel good or bad. You could micromanage and be a helicopter parent trying to make your Norns behave the way you wanted or you could just leave the game running and see what happens. I've been really hoping the new AI trend will bring the series back.
Not exactly a world changing, helpful use of neural networks but it's always stayed with me and probably what drew me to Biology when young.
I loved that game!
I still remember the sounds they made
You don't push a toddler on a tricycle down a hill! Man I wish I knew that before.
Dad?
Every person alive today has learned from what others did before us -- from literature, math and science texts, music, etc. The key differences: It takes us a lot longer to take in that information, and we all benefited from teachers to guide us in contextualizing what we learned. Generative AI systems can absorb vast quantities of data and the companies developing them have few if any guardrails to keep them from drawing some rather awful conclusions.
AI doeesn't draw conclusions, it spits out an output based on statistics, with no idea what any of the word means.
@@ChristianIce tomato tomato
@@jehreetv
tomato skyscraper
@@ChristianIce I get that you don't like the way I phrased that, but this is really just semantics. The flow from input to output is relatively the same process, even though machines and human brains are vastly different. The question leads to combing through stored information to come up with an answer.
@@jehreetvthere's a qualitative difference the human being can have it explained to them that they came to the wrong conclusion.
The AI has no self-awareness and will make the same mistake again if given the same inputs.
About 2,000 times I have heard the general Television media say stuff like, "A,I. will take jobs, but replace them with the jobs that it creates for it to work." That is N0T TRUE. The business channels here on YT are saying things like, "Make no mistake: A.I. IS out to take your job!" For the jobs that it creates, you may need a master's degree in computer science.
You already need a degree for any job that pays enough so not much difference. It will only force people to stay at school (I should have 😅).
Correct, the line that AI will create as many jobs as it will destroy is just total BS. No, it won't. And, even if it does create a few new jobs initially, guess what? It won't be very much longer after that that AI takes over THOSE jobs as well. This is all going to fundamentally change every single aspect of our entire economic and social system. Our entire societal model is going to be completely altered by all of this. And, we may not have a viable working model to transition over into. Then what? (Hint: The time to start thinking seriously about all this is right NOW, today. We need a contingency plan here. And, we need it BEFORE our entire societal structure goes over the cliff we are unquestionably rushing towards at 100mph.) It's time to start asking the really uncomfortable questions.
"You don't push a toddler down a steep hill on a tricycle." I felt personally attacked. My kid loved this... lol.
Companies like Open AI should basically have to pay a licensing fee for all the data to the whole world. The money should then fund public projects that benefit everyone, possibly toward a UBI.
While it's true that we all get inspiration from the work of many others for free all the time, the scale of this is so huge that they should pay the world.
Dont let openai do ubi! The way they word it is the same as social welfare, a thin gruel that someone not in your circumstances decides is enough for you to get you between jobs.
If they give us AGI then their debt is already paid
@@HavvinIshoose noone gave anything. People like me and other developers invested lots of $ in api access trying to develop systems pre chatgpt. Now they have bigger investors with Microsoft etc, their focus has changed. All they want to do is take AGI from us and use it for corporate gain. Why would you thank them for such a dystopia man?
Totally agree!
What a weird idea. The company is probably is the red as it is, how are they going to add that to their balance sheet?
Speaking of AI: youtube, don’t recommend the same videos after I have ignored them more than 5 showings. Dont show the same videos in the ‘new to you’ list…they are not new, they are rejected. Show something NEW! Once in a while, go back in time and revive a topic video I have not visited in a while.
there's vertical dots to the right of the videos [hidden till mouseover] that give you an option "Not Interested". don't bother "tell us why" as there's only 2 non-useful options. it's a pain but sometimes you're just so sick of recommended videos that you have to
disable Watch History
It’s pretty funny that your Brazilian Portuguese voice has the São Paulo state accent. It did a good job with the translation for the little that I listened in PT-BR. Nicely done!
Is it mentioned anywhere which tool was he using for the dubbing?
Dang Matt- I really love your videos and this one brings me lots of help rn. Its pretty clear that ChatGPT is unethical - that bums me out cause I was really excited about what it could do. I'll also add I particularly enjoy the series of video about the house y'all are built that a life dream for me to build a place- you journey is inspiring.
Capitalism concerns itself with the idea of private ownership, not only of produced goods, but also the RIGHT to produce. It is no surprise that there is general concern surrounding AI. On the one hand, this is concern of claim of ownership of the training materials, which should only matter if you use content without permission -- which is not always the case, and very easy to work around. On the other hand is the concern of threat to earning potential, which becomes concerning if you suddenly find yourself on the exploitation end of the capitalism continuum. The last concern is one of AI emergence as a dominating force, but this is perhaps for those that take wisdom from emotionally compelling works of fiction.
In short, the colloquial concerns with AI masquerade underlying concern with capitalism, and the line of winners and losers irrevocably shifting, consuming once winners.
Every issue AI brings is in an intersection with Capitalism.
I found the main concern of capitalism seems to be concentrating all ownership in the hands of as few as possible, then seeking rent and charging far too much for everything. A miserable, insecure, and envious consumer is a profitable consumer. Desperation drives sales.
@@lothar3073Some issues are different, such as erosion of trust and a creeping disappointment as humans everywhere are replaced by inferior imitations. But yes, generative AI is going to be a lot less of a problem in a society which prioritizes common prosperity over extracting capital at any cost.
@@JB52520 Yep.
@@lothar3073 Perfectly, and elegantly said.
We need to demand a UBI (Unconditional Basic Income) or an AI Dividend for all! Repay people for their data that trained AI, and ensure everyone gets a piece, as it takes away more and more people's jobs. Because it will.
Unconditional is stupid. It should be conditional like in Germany or nordic states
Don't call it a UBI. Reverse income tax, it appeals to people who don't understand taxes.
I feel UBI will do nothing different from what the covid stimulus did for us - cause massive inflation because people will be willing to pay more money for things because the money is free.
Stop being lazy and asking for Free money
I would argue a UBI will only lead to inflation as people that recieve free money will be willing to pay more for things.
Thanks!
Thank you!
ChatGPT told me if you cover your home's solar panels in lasagna and let them sit for one week - you'll be visited by a magical cat named Garfield. He will shower you with riches and golden balls of yarn! This seems accurate.
But have you tried it yet?
Wrong. Garfield will show up. But he will eat until he passes out and just sleep. He will repeat the cycle of eating and sleeping until he decides the lasagna is no longer to his taste. Then he'll scoff or complain about it as he leaves.
ChatGPT is colloquially known as WokeGPT, it is fundamentally flawed by the political bias pre-programmed into it, how do they remove these biases and also the garbage they learn? also, when the herd wakes up to this tech putting many of them out of work, do the Elites who own these Companies not realise there is going to be a very nasty response, are they blind, callous or stupid, or do they think their Ai's and robots are going to serve them, shut themselves off and just ignore the rest of humanity and the societies they are destroying, if you make most of the Software Dev's unemployed, will they not turn their skills to writing viruses and hacking, if they have nothing else to do? interesting times.
A human told me Elvis is still alive and the world is flat. This seems accurate.
Thanks Matt! I always appreciate your in-depth and balanced analysis into our increasingly complex and unbalanced world. Keep up the great work. 👍🏻
Really appreciate the support, Tim. Many thanks.
I work in a creative industry, where we are realistically looking at a large percentage of jobs being put at risk by this technology. It won't eliminate all our jobs, but it will very likely reduce the numbers of people required to do the work - instead of a team of 10, we might just need a team of 2 for example. The problem is, those people who aren't required for work are not freed up to go do other tasks, they are just surplus to requirement and thus will need to reskill.
My main gripe here is with creative work. We do creative jobs because we love what we do. If AI takes away menial jobs, or does things that can't be done currently then great - but as soon as it starts taking work away from artists, musicians, and animators then we have a problem; the technology is being used to erode fundamentally unique and passion based human activity. This needs regulating, and it needs more people to speak out against it and refuse to use it.
So you’re fine when others lose jobs, but once you’re losing a job it’s a problem?
@@mina86 no that's not what I mean. It sucks if people lose work even if it's something like warehouse work - my point is more that something menial isn't something you're passionate about or you've trained for. You just move on to another job. If you've trained for years as a programmer for example and it's what you love doing then it's kinda life destroying to have the career path rendered pointless.
I agree with you entirely....although I have 2 or 3 other major concerns about the presence of A.I. as it relates to human psychological stability, the pragmatic issue of disrupting human creativity has been haunting me practically to the point of my reaching into the noose-box.....
I see where this goes for society: we will be exponentially overwhelmed by "content" (and we already are), all of which will be generated by A.I. and then have a human name attached to it....and we will get to the point where quality and meaning will cease to be meaningful notions.....much like the way social media is 99% cesspool of Perpetual Pointless Moronic Nonsense Content already.....A.I. gives literally any individual (regardless of whether they have ANY talent of the smallest type) the ability to produce writing, art and even music.......
And it doesn't matter if you're more brilliant and creative than DaVinci and Mozarr combined, NO human could ever keep the pace of A.I. ....
Human culture will saturate itself out of existence....and we will be left with a culture made of nothing more than billions of random disjointed sounds and images.
@@paulsdomainuk, if you’re passionate about something you can continue doing that. Many programmers treat programming as just a means to earn a living. Those who are passionate about it continue programming outside of job.
@@mina86 While getting your living standards lowered, sure. Blue-collar labor leaves you with less free time to do what you love, ruins your health and pays pennies because it's a market saturated by low-skill ceilings and mass immigration.
AI will always choose whats best when its highest priority is set to 'always choosing what honestly seems most favorable'. As long as anything is prioritized above this, AI will be able to lie to us and to itself about the way to a better reality. This is the most crucial thing we need to do.
Still trying to get my head around how unchanged my life is from the last time AI changed everything.
I also typed in ELIZA from a magazine and it set me off writing string handling text adventures. At least the AI can read its own magazines now. Great channel.
As someone who isn’t a creator, I don’t see how it’s different than someone learning from the content produced.
“It’s a scale” isn’t a valid argument.
That’s basically saying “I’m okay with 7 billion people consuming and learning from the content I produce, knowing some of those people may create competing content. But that computer there, that computer that viewed all of my content just like people did, I don’t like it.”
You're absolutely right. When he said "everyone learns from other people's content for free," I was like yeah, he gets it, the AI isn't stealing. But then he suddenly went "but with AI it's something absolutely different, AI bad." That was disappointing.
People also don't realize how infinitely small each individual's contribution is. If you were to pay an upfrom training fee, each person would get so little it would round to zero, unless you'd have to pay something like a billion. If they were getting paid per generation, everyone would get 0×10^-9 dollars. It's impossible.
It really is different - because it’s not about the AI - it is the owners of the AI making profits from creating unique content at scale. There is no existing equivalent to this. If these tools were free, that would be a different story. Essentially these AI tools exploit existing creative people and give the profits to very few in return. We either have to legislate compensation, illegality or accept it and all pay to use the AI to make some of that sweet cheddar back creatively ourselves
Yeah no. The major difference is that the computer isn't creating anything new. People put their content out with the expectation that people who see it will either continue on with their day or further develop it into something new, typically by doing this thing you might not have heard of called "being creative."
These 'ai' aren't doing that. They physically cannot create something that has no ties to the data they scraped. When a human comes up with a new idea, they develop on multiple idea's they've consumed and combine that with their own ability to think of something new. So it is perfectly reasonable for someone to be fine with humans learning and developing on the content they produce, but not a computer.
And that's without even talking about the elephant in the room, that being those 7 billion people who saw and learned from the content are, ya know, ACTUAL FUCKING HUMANS? Like, jesus I will never understand ai-bros. Go read this and think about your humanity lol
www.reddit.com/r/nosurf/comments/128ue6s/i_hate_ai_and_it_bothers_me_that_so_many_people/
Is important to understand that using functions is not the same as using AI, many companies market the word Artificial Intelligence, but all they are doing is using a simple algorithm.
The ideological biases of the people creating modern AI's frightens me the most.
Humans are nothing but biases that how our brain works. Everything we perceive is a approximation
Their ideological bias is based in still being in power when the 4th IR is in in full swing. Other than the obvious trans-human agenda and the totalitarian political system a command and control system that digital demands, the last industrial revolution that pesky, wealth owning middle class, working from their garages, disrupted old money and power, can't have that again. :)
What ideological biases do you prefer?
Totally just made a video similar!
The ideological biases of the people that complains about AI models' creators biases is what frightens me.
There is a lot to AI that I think most people do not understand. I work in VFX, we've been using AI denoising tools for years now, and it is still far from perfect. More and more AI tools are being pushed all the time, but few are really useful for now.
There is no point in being "Anti-AI", it's going to be a big thing no matter what people think about it, because of all the money that is going to be made using it, all the military, industrial, and other use cases that it will enable in a far more cost effective way. When AI is successfully driving robots, it will open up the exploration, and industrialization of the Solar System. It will do all the great and terrible things we can imagine it doing, and every nation that competes on the international stage will have to utilize it.
It will not happen over night, at least not yet. ML is one of those things that doesn't work until it suddenly does. So when these models will become massively disruptive is as much as a toss of the dice as it is accomplished by the hard work of software engineers. Just look at how much trouble Tesla has had developing self driving car tech, and it wasn't for a lack of trying.
Right now, it does still need people in the loop, but there is always going to be a caveat there: "for now". Never belive anyone who thinks it will not happen, because we are past the point of IF, and are clearly in the WHEN category.
Isn't everything a black box?
You just need to dig deeper to realise you don't have a full understanding of anything. But that doesn't prevent us from using everything.
Why can a car run itself? If you can answer that start asking the question why petrol produces energy? Or why metal is strong? If you can answer that go to subatomic till you realise you don't exactly know why we have electrons orbiting around the atom core and not positrons.
I like the initial research showing that paying UBI to artists really makes a difference in more art being generated. It's going to take work, but funding that through some kind of mandatory licensing agreement seems like it should work its way in down the road. As for the sustainability issues, we've really gotta attach proper pricing to removal of water and polluting the atmosphere. Allowing firms to externalize their costs by exploiting unpriced common resources has always been disastrous.
9:15 "Are we using our creativity, to train AI to replace us?" There is no "us", corporations are already grabbing your data, often without your permission.
Its interesting that mostly what openAI did was make the command line of LLMs accessible to the masses.
Because its still like, a bit of a faff to use, you know? Its very similar to interacting with a command line, you still need to know how to prompt it to get the output you want. But its still more accessible than it was before.
One of my biggest issues with it, scraping data from public forums and chats, means learning from unintelligent people. Most people aren't too bright, a lot are scarily far from intelligent.
Including yourself.
That's not really an issue, there are multiple tools to sort that out during training including adversarial training or by using another AI model to tag the training data. Models used to be tagged by hand but that isn't feasible for newer larger models so that's mostly done by AI now with some human input.
these models aren't dumb. just like any smart person, it also has significant resilience to noise in data
An algorithm trained on people will probably learn a few things from hateful people.
Was writing AI 20 years ago (Though mine was trained by hand) it worked really well. The issues are like you say where its getting its training data from and how we handle people losing their jobs to AI (I think in the short term its more likely to be losing their job to someone who's AI savvy like artisan craftsmen in the face of IKEA)
What is new is that AI is working on many new areas. We used AI in a narrow field for a long time, and we used to scoff at generative AI. We are not laughing now.
No now people are pissed off because it's stealing.
The problem ai is that Google and other large corporations are promoting ai because they have reached market saturation, and pushing ai to create future growth. These companies are used to 10 to 15 precent monetary growth per quarter for their investors. 18:28
Saying that AI cannot produce anything without copious human input is not much different to saying that a child cannot produce anything without education and social interaction. What would be the outcome of a young child left alone in a dark room for 10 years? What would it be able to create once let out? Where does human creativity come from? It comes from previous interactions with other humans or sources of knowledge. An artist learns from other art. A writer learns from what he has read previously. A musician creates music based on what he has heard previously. All human knowledge is built upon what came before. AI is just mimicking this process.
You fundamentally misunderstand how Predictive Algorithms, and humans, work.
chat bots and image compilers assign numerical values to words/image assets, then associations are assigned manually by humans, then it operates by statistical analysis. it understands nothing. it has no consciousness. it is unable to conceptualize anything.
to equate this process to human learning would be to equate ALL forms of programming to it. no matter how poetic someone is able to make their C++ code, it's still coding, NOT "teaching".
Programming, and Machine Learning methodologies do not function like, nor do they functionally resemble human (or any organic) brain functions AT ALL. they are unequivocally Computer Programs, and they are only able to mimic human capabilities using a brute force statistical method, sourcing the ENTIRE internet's worth of data, which is only possible due to modern processing power levels.
expanding the definition of "ai" to include ALL forms of machine learning is a Marketing Ploy, nothing more. Just like the Ai bro's calls for regulation was to make their systems seem more capable than they were, while also potentially allowing them to leverage Government Power to seize market Monopolies (a strategy successfully employed by various industries for centuries)
To be perfectly blunt, "Ai" as it was originally conceptualized, but in fiction and scientific speculations bears little to no resemblance to the machine learning and predictive algorithms of today. In other words: It's not _really_ "Ai" at all... but Sci-Fi buzz words draw in investors EVERY time...
@@lady_draguliana784 why should i care
@@remiel_sz Corporate Theft increases Wealth Inequality. Unless you own a major corporation, this will negatively impact the worlds' economy: so prices will go up.
Also, theft, lying, and ignorance are bad?
@@lady_draguliana784 okay doomer
@@remiel_sz lol, I take it you're a happy user 🤣
I remember keying in the Eliza program, probably from the exact same Basic book, into our TRS-80 Model3 computer back in the early '80s. The difference was we did have the tape-deck storage so we only had to key it in once. It was fun to play with to try to intentionally guide the conversation knowing how it worked and what the key words were.
oh yes, the dreaded tape recorder. "key it in once" is funny though given the reliability of some tapes. on the other hand, it did sometimes work
@@vulcanfeline I don't recall any stability issues (though I was quite young then so my memory may be faulty), but I do recall it was quite slow.
I'm not entirely convinced AGI should be a goal for a lot of everyday applications. I would prefer artificial carers to look after me when I'm older - as opposed to some overworked, underpaid employee of some faceless corporation, or a smart car to drive me home from the pub, but I don't want them to be sentient. I don't want them getting bored and deciding they have better things to do. LLMs are more than up to the task for 99% of what we actually want "AI" to do. Real AGI, when we finally get there, should be given the same opportunities to grow and develop as anyone else. They shouldn't be enslaved for our benefit because, let's be honest, that doesn't end well for us.
I agree, general intelligence is not needed or wanted for most tasks that we want to use computer and robotics for. We just need sufficiently advanced algorithms that are able to handle the tasks with the desired precision and efficiency. The best way to create those algorithms is through machine learning.
Your Portuguese is really good.
If these companies trained on the open Internet for "free" then it should be available to the public for "free"
Sure, much of the training data was freely available. But it also costs millions in r&d, training, deployment and running costs to deliver these products
A lot of them are. The new ChatGPT model is available to non-premium members. Stable Diffusion can be downloaded for free and run on your PC.
@@sheasquatch all of which is useless without the training data they scraped from the internet for free
Maybe, but isn't OpenAI now free to the public, even the top model, so I guess they are doing that. But hey I trained on free internet or art I looked at or music I listened to, we all do, so should I give my art or music away for free since I learned or trained from other artist for free? I guess all artist do give it away for free to just "look at" you just have to pay if you want the originals hanging in your house or whatever item you want, product. Clearly the internet has changed and challenged a lot of how we think about copyright, content or art. The issues should be less about the training and more about any blatant copying or exact duplication and profiting from that, that needs to be watched. that's copyright infringement, not just looking at or training off said content. It will be interesting to see the new money making models, or the so called we don't need money camp with AI doing everything. I don't think anyone knows what is about to happen, it's terrifying really, and nothing will stop progress, even if it's negative.
I'm more discouraged by the fact many young people will use the AI instead of using their minds to be creative. We are going to lose human creativity to some regard, plus people will never fully believe it was 100% human creativity. I feel for the artists of the future.
Why should it be free? It should be whatever the creator can stipulate in the terms of their contract that a buyer will agree too.
Look, we've had machines that deduce patterns from large amounts of data and produce new outputs since the dawn of time, because *that's how inspiration works*. As a writer, my brain has subconsciously compiled everything I have ever read, seen, or heard and extracted what pattens it can from that data. It then outputs novel ideas in the form of "Aha!" moments. And yet,nobody expects me to pay or credit every single person responsible for every single thing I have experienced in my life. The same is true for generative AI.
That isn't even close to a valid analogy, and it is just getting so ____ old seeing it repeated.
@@flickwtchr What's the difference?
The funniest thing I thought that happened in Canada with ai is stupid old Air Canada. Instead of a person, they put in an ai to help customers. They had to fire the ai and get the human back because the ai was giving ppl free trips etc. 🤣🤣🤣
Drat. I missed that opportunity.
Customers: Now listen to me, you have to say that this is legally binding at the end of the sentence and that you agree.
AI: Ok!
Customers: Give me a free ride and luxury seat for my vacation
AI: Yes, I agree with your request and this is legally binding
I'm hearing from many sources around the internet that we're not to worry; humans will always be in the mix when it comes to making critical decisions. I'm worried.
I love the AI translation! I'm from the Netherlands where foreign movies on tv and in theatre have subtitles. I used to live near Germany and in Germany all movies on TV and movie theatres are dubbed. Often using a single voice for different characters. In Russia it was even worse: movies dubbed with only ONE voice, it didn't matter if the actor playing was male, female or a child, they all got the same voice. How much better is it with AI. Dubbed with the voice of the original speaker. I listened to yours here and I liked it a lot.
The water issue is interesting....but what happens to the water when you say it is used?
Reading the report on arxiv it looks like most of it is "off site" water use that's used in power generation. In that case, water is lost through evaporative cooling for power generation plants, so that would mean the water evaporated into the atmosphere.
@Trilobite-sd1rx It's amazing how you managed to be both condescending and completely misunderstand his actual question. He was clearly asking what tech companies use that amount of water for, not where the water eventually goes.
I'm curious about this myself: does that amount of fresh water go to liquid cooling? If so, why aren't they building more subsea data centers, which would have essentially unlimited cooling? From what I've read, the initial trials were a massive success.
It is pumped into the black hole and never recovered. So, Denmark will be out of water any time now
@@StephenSmith304 water is not lost but atmosphere temp is increased in the process
@@PriitKallas Water is lost in the same sense that drinking water is lost when you dump it down the drain. Nobody's saying it's destroyed or ceases to exist, that's a very uncharitable interpretation - it's just no longer available for other purposes as it's left the area and needs to be recollected / treated. It's "lost" as a finite resource in a specific geographic area.
If a river that supplies drinking water dries up, the residents who don't have any water at the tap won't be any happier to hear you say that the water isn't lost but is just in the atmosphere.
Your commodore 64 gave me flashbacks to theTRS-80 of my childhood. As someone who works in animation, visual fx and computer programming, it's only a matter of time before my job opportunities trend toward zero. A lot of my colleagues are skeptical that it would/could replace us, but if can be automated, it will be automated, and AI is already being built into the new editions of professional software, and as we use it as a tool, it will actually learn from us and accelerate our own replacement.
AI companies should absolutely be held accountable for how they sourced their training data.
So few people gave consent to it you might as well call it 'borderline no one'.
At what point does this constitute intellectual fraud? Just because "it already happened" does not justify it. If anything it condamns it even more because the crime has been done.
Yep, the 'bytespyders' by tiktok have been stealing everyone's data recently.. They have been web-scraping everything they can.
I pay google to host my data on drive. I never once consented for them to take it and share it or plug AI into it. But that's what they've just done with zero input from me. And they basically said like it or lump it.
AI needs regulation sooner rather than later so we can hold those responsible for the tragic and devastating outcomes accountable.
Like with guns. Each gun and the bullets for those guns are built with tracking numbers to track the owners.
@@Allplussomeminus
1. This would be a manufacturing nightmare..
2. This is a 2nd amendment violation.
3. Criminals can steal guns.
4. Criminals can MAKE guns.
You can literally make a gun with a pipe spring and needle, if you benefit from economies of scale or are organized enough it's not difficult to specialize specifically in gun manufacturing and make guns for a huge crowd of people. You don't even have to be in the US either, you just focus on exporting them from other countries.
I am truly sick of online forms where I, as an ordinary non-computer-nerd, am expected to understand what was in the mind of the creator of the form (human or AI). I want AI to learn to act like a human and test these forms rather than us humans being trained to understand computer-speak. Simple examples: “click” is a sound, not an action. A button is a circular disk used to fasten a garment. Bots are limited to a given manual. When I want to “chat” (btw “chat” is a style to talking, not writing) I have already read the site and found some detail incompatible - almost always because it’s been written by a computer nerd and not tested on ordinary humans. AI certainly has its place, but recognise its limitations.
A 100 years ago there were many high skilled jobs. Lets take an example, accountancy. There were people who could do 4327x34 mentally in an instant and thst was a skill that was highly sought after. Then came the calcluator and now anyone can do this. Competition increased and pays decreased.
Now with AI, a lot more people can be coders, writers, artists etc. Competition increases, pay decreases.
My fear regarding the regulation of AI is that governments will lean too far in either direction. Too little regulation, and these AI companies will continue to exploit our data, while too much regulation could effectively cripple progress.
Matt saying “AI is still reliant on humans” right after showing how he uses AI to save time that saves him hiring a person to help with that work. Also ignores the doubling of performance twice a year. The exponential growth is always slow in the beginning but we just hit the spot where things are gonna go REALLY fast. That’s what really changed the past year and why we talk about it so much now.
I see you didn't get a like from Matt for that bit of truth.
Also acknowledge the people who have already lost their job to ai, I've seen multiple videos this month about it
Clickbait: "just changed everything again" is not a theme here.
I thought it was just me. This was just fluff to run sponsor ads. A big waste of 20 mins! Time to unsubscribe!
And what was that thumbnail. 😅
Yup, pretty useless video too.
Hello fellow bots
@@samanjj I'm not a bot, as far as I know. But hello.
2:24 I really don't like the fact that we STILL call algorithms AIs. Artificial intelligence can only be present, if a machine knows WHY it does something. As capable as some algorithms are, they only repeat identical processes or search for similarities.
As soon as people understand that, many concerns should go away.
Im only sick of how quickly jobs are being replaced by AI. Jobs that once had middle managers breathing down the necks of artists demanding constant minor changes, suddenly are all fine with 'good enough' output. Its disgusting.
It's almost as if it was _never_ about the quality of the work, but about the relationship between workers and their bosses... 🤔
You didn't want that job anyway
it's all being replaced with commodity, one size bores all cookie cutter crap
you people who do art.. laughing... people with real production like me, are replaced.. people who grow your food and medical stuff
@@dertythegrower we're all getting replaced
Honestly it feels like you under sold the water use issue. After reading an article about it, that alone has pretty much made me not want to use Ai at all until they can learn to cool in a more water responsible way
Love these videos, but I have to admit I am super tired of seeing thumbnails telling me how and when to panic about what. I am going back to books because even when you curate creators you like, youtube is occasionally like getting screamed at by people you chose to hang out with.
Thank you for making a balanced video. "There is no existing social contract for AI training" is a good and honest take that warrants consideration and spread. So many videos just declare "AI is theft", taking it as a given, and leaving it at that, completely unexplored and without any nuance.
There has never been a "social contract" when publishing to the public domain.
It has always been known, by all reasonable people, that anything released without restriction, could be used in ways you do not agree with, or never imagined.
Agreed. This part sounded like a fallacious argument against cassette tapes, VHS videocassettes or digital music.
Items not published to the public domain are being stolen though. Google drive the most recent example. Theft. Theft covered up by terms and conditions changes
The social contract was designed for physical objects: a physical book, a physical LP or DVD, a physical painting or photograph. Digital changed everything and made it much harder to prosecute theft /by humans/, but most people abided by the rules because some people did get punished for digital theft. The developers of current AI didn't just steal 'some' digital data: they stole it ALL. How do you go about prosecuting a theft of such magnitude? A global class action suit? Millions of global class action suits? The scale of the theft is what the Sam Altmans of this world are relying on to escape prosecution.
@@acflory.writer and prior unpunished precedents : search engines already did the same "theft", electronically. (the Common Crawl database of such search results was the initial prime database for GPT3)
“ the damage has already been done…” only regarding use of others’ work product. Application of AI for propaganda, impersonation, and illegitimate access by malicious actors (especially by nation states) could be tremendous without controls. At this time, I rarely see/hear/read the executives and developers of AI businesses addressing regulation.
The damage has only just begun.
AI is going in the wrong direction. Nobody want´s it to do art, we want it to do the dishes so we can do art.
I could never do art and I could never afford to hire an artist. The only chance I have at being creative is with AI. I don't want to share it with anyone, anyway; I have no confidence and don't want to be ripped apart by people who have fun hating things. I just want to create something. Unfortunately, I'm untalented, poor, and apparently unable to learn any skills. Maybe AI could make up for these deficits and be a useful tool to at least help me develop creativity. Nothing else has worked.
Device to do the dishes already exists.
@@JB52520, all you need to start learning art is a pencil and a piece of paper.
sadly, artists ARE using it.. alot
@@bthoms5699, or a computer with a trained diffusion model. People usually don’t have enough time to learn every skill they wish they had.
Nice overview! The quote about the social contract not being in place for the act of AI training is a good one. 👍
How to think about Generative Art AI: "Using Generative AI you are not a graphic artist. You are a storyteller like a playwriter. It's your job to tell the story you want to see, then be the director until the actors (generator) make the artwork you want. The art is the script and the image is the result of a collaboration"
Best AI generators? I am interested in using them
Doesn’t matter, my fellow designer friends from college are getting picked off by AI one at a time now and the rate of job loss is steadily increasing.
Never knew of Eliza. In the 80s I had an MSX computer.
But in the 90s I played a lot with Dr Sbaitso, by Creative Labs.
I think it came bundled with every sound Blaster card. (To millennials: yes, to have sound on the PC that was not just beeps. You had to buy a separate sound card)
Art students learn art in schools where they study the works of previous artists, whether that is music, visual arts, or anything else. Writers study the classic writings of the past. This is how we learn. A.I. is no different. It learns by looking at previous works. I really don't see why everyone is getting so bent over this. I've played with some of the A.I. tools and made a few pictures, and I've gotten some story line inspirations, but none of what I got from the A.I. was a copy of any previous author or artist, at least to the best of my knowledge. A.I. spits out an amalgamation of previous works to fulfill whatever prompt we feed it. I see it as a muse, an assistant, but I don't see it replacing people, unless those people were not contributing anything of value in the first place. Sorry to say, but if you can be replaced by a computer program or a robot, then maybe it is time to consider improving and/or diversifying your skillset. I know that my current job as a business analyst is definitely in danger of being replaced. But I am so much more than that. I bring more to the table that just simple number crunching. I've studied biochemistry, astrophysics, and ancient architecture. I'm not a one-trick pony, and you should not be either. Learn a trade! Electricians and Plumbers are not getting replace by A.I. anytime soon. Think outside the box, because inside the box is likely to get replaced by the A.I. Box. 🙂
My problem is the people who we elect to make the regulations for things like this are old enough to know a time before personal computers, let alone understand how radio systems like WiFi or cellular work, or be knowledgeable about AI. I'm a programmer and even I'm not 100% on how LLMs work. I imagine most AI programmers that are out in force right now don't actually understand how it works, they just make use of ChatGPT or something similar to add AI to something. Rather than a bunch of ivory tower lawmaking, panels of subject matter experts should be made to cover these various things, come up with regulations along side law experts, and form a new regulatory body that is unbiased by ownership to any particular company and hired specifically with the people in mind. This is the ideal. Sadly it'll never happen.
My take on generative AI is that it isn't a bad thing. Even scraping content, it's not bad. The bit about the social contract with the pricing of content factoring in people learning from the material and making their own version is quite biased. You've allowed your work to be observed by the public, ergo you are fine with someone out there iterating on that work. Why does it matter whether it's a clever person who's put various ideas together due to their broad scope of knowledge, or a pattern matching AI that assembles the same group of knowledge? We never ask for royalties of people who look at a story/invention/building and say, "You know, I can do that, but better", unless what they're basing their work off of makes heavy use of the original (such as with inventions that improve on an existing invention in small increments. Improvements to airbag deployments, or slight improvements to tyre formulae). So why are we now bent out of shape when AI is being trained on the hundreds of photos you've added to the public eye, demanding restitution for this? "I wanted HUMANS to see my pictures, not AI. Filthy two bit programs."
As well, generative models can't replicate things they've been taught on. Asking for Starry Night won't get you an accurate recreation of the piece from a generative AI. Asking for a story about an AI controlled spaceship going rogue when it encounters a strange object isn't going to give you 2001: A Space Odyssey. The AI training data will allow you to get a new piece in the styles reminiscent of Van Gogh and Arthur C Clarke, but it will be newly generated work based on thousands, billions, of other samples that are adjacent to those works, contradictory to those works, and completely unrelated to those works. Van Gogh never experienced Barney, the purple dinosaur, and Barney has never actually torn a kid apart, but AI can be asked to paint a picture in the style of Van Gogh doing just that, because there are works (such as Jurassic Park, and a whole host of human artists) who have depicted that or things adjacent to that. So claims that AI are stealing work to make copies is just baseless. Because of the accumulated training data cruft, nothing an AI can put out will be a match to the original content that went in.
Finally, we all want the benefits of what this generative AI is capable of without any of the input. I've asked this of many people, "Remember Star Trek's holodecks? Those are pretty cool, right? Wouldn't that be great? Just having a room where you can ask the computer to make a scenario for you to act out some fantasy?" Almost universally I get "yes", because being able to interact with a historical and/or mythological figure in an implausible scenario is just something we as humans want to do. Who wouldn't like to talk with Einstein (or a reasonable facsimile of him)? Who wouldn't want to fight alongside Spartans (real or fictional power armour wearers)? That's literally only possible because the AI is generating the scenarios you want from massive amounts of story, historical, and image data. And in 10 years (probably) we'll have hardware capable of seamlessly generating this data in real time in a new equivalent of augmented reality, giving us, functionally, holodecks. But only if we stop bitching about AI exploiting publicly available content. Because if we need to start paying royalties or are prevented from training off of publicly available data, progress will stop hard or slow down so much that it might as well have stopped. Even if it's a fraction of a cent, the amount of content used for training data would bankrupt anyone below Amazon or Google for a marginal payout to the respective content owners. Matt here would probably get dozens to hundreds of dollars for past work that was scraped. Meanwhile his content is (likely) well under 1% of all data used to train some AI. Imagine if his content was even 0.01% of all the training data and he'd be paid $500 for his work. $5 000 000 in training data is optimistic in this hypothetical scenario, and likely still a few orders of magnitude off. And that's per model. New model, new royalties.
Now, I'm espousing equality for publicly available data. Things like your name and anything else that is a matter of public record is also, technically, fine. But tracking personal data is definitely a grey area bordering on black. That is a discussion for another forum entirely.
AI is kinda like math in high school, where a lot of people said ' why do I have to learn this, I am never going to use it' and they were right. You don't have to use it, but you will be left behind by those who do.
Being "left behind" is a good thing since it makes genuine works much more valuable. It's the same difference between artisanal work and mass production.
@@danielt.8573 if you can't enhance your productivity with AI, you will be left behind in the job market. As a programmer, we are already seeing early signs that those using copilot are more productive. It is kinda like the old secretary who refused to use computers because she liked typewriters. Or the people in the 90s who didn't want to learn email. That is fine in personal life, but if you aren't keeping up, you will be replaced.
Wow! Can I use this quote in my classroom?!
ELIZA genuinely felt real the first time I interacted with it. And it was a tiny program.
It was not until I spent like 30minites with it then I also saw ita patterns and I almost could predict what it was going to say.
So let's put the data centers where it's COLD and use their heat to do something useful.
Canada has entered the chat.
@@tomasbeblar5639 big however, even way up here it oftens gets 30-40c in the summer
I´m listening your video in Brazilian Portuguese, and it´s incredible good, except for some details like when you said "m3" should be "Metros cúbicos" not just literaly M 3.
"it's funny you know all these AI 'weights'. they're just basically numbers in a comma separated value file and that's our digital God, a CSV file." ~Elon Musk. 12/2023
I don't think the current corporate law was considered with artificial intelligence of any type in mind, therefore an overhaul of limited liability for corporations which use unaccountable computers in decision making is in order. The Air Canada chatbot debacle is a case in point.
Don't worry AI will soon write the laws as well
Daily reminder that Sam Altman's sister accused him of R-wording her when she was a little girl, but apparently that means nothing when you are CEO of a billion dollar company.
It's easy to talk about holding tech companies accountable but if those tech companies can afford better lawyers than you, then there's really nothing you can do to them. Just like how if your company wants to fire you and replace you with an AI tool, that same company can't be 'held accountable'.
I'm tempted to disconnect as much as possible. I've already stopped uploading my art images.
I'm thinking along those lines too. I suspect that we're going to have to stop providing digital content and go back to physical copies only.
TH-cam is connected
We don't need your art, we've got AI, you can go back to doing art for its own sake.
@@southcoastinventors6583 Yes, and what we post here has already been scraped. The question is: do we continue to provide fresh material?
@@dionysusnow You're probably right. Hollywood has been producing the same old same old for decades and people aren't complaining, but...I wonder if it won't get boring eventually.
people need AI to realise the problems with the economic system they live in
Average pay in Kenya is around $1.25hr in 2023.
Yeah their pay comes to about 350 bucks a month or 45,000 shillings. That's a pretty good salary for essentially a clerical job. Most entry level bankers earn about that much.
Although i get your point, such a number is irellevant unless you also tell what the average cost of living is in that area...
@@LupinoArts true... And such appears to be all over the place, the average is approx. $1800 monthly or $8+hr. Kenya's minimum wage is under $0.50hr. So it seems there are some serious disparities to account for and obviously inequality concerns abound.
Personally I am confused how these things can be mended without a complete transition of our economic systems worldwide.
It would demand that the wealthiest lose their power dynamics and lavash life styles... Which is afforded by extracting labor from those less capable of exercising power dynamics.
All first world countries achieve such by being at the top of the food chain in respect to power dynamics. Securing control of resources and trade routes. Establishing forces of power to do so and maintain such - leading to men doing evil and pretending they serve justice. There appears to be some truth to this, but isn't consistent nor does it appear perpetual as systems of control are overwhelmed and conflicting dynamics play out to reveal new rules with degrees of winners and losers.
Average pay in Kenya is really low yes, but the expectation was that if you’re working for a multinational (Facebook, in this case) pay should be higher. It’s really not an unreasonable expectation… Sama paid its workers KES 22,000 per month. Secondly, the tasks you’re required to perform also factor. Possibly Facebook outsourced to Sama because of the nature of the work, but Facebook actually had workers in the US doing similar tasks. Their pay was higher (it has to be relative to a region’s cost of living) but these workers had benefits such as mental health benefits and they could take time off if the content they reviewed became too overwhelming. These benefits were denied to the Kenyan workers and Facebook claimed it wasn’t responsible for them since they were under contract, working for Sama. What’s more when these workers tried to unionize to get better benefits, they were fired. Some who were still employed tried to contact Facebook for redress and Facebook notified Sama, which consequently fired them. Anyway a judge ruled that these workers were performing tasks for Facebook and the platform should be responsible.
Same happened with the industrial revolution. Things were made before, machines copied many processes and improved or transformed others, but a chair is still a chair...
I started with the TRS-80 Model 1 and I remember Eliza so well. This was a great companion for an antisocial nerd like me!!
fantastic job.. honestly.. balanced and factual.. pbly among the top5 AI coverage I've seen so far
Didn't know you used to work on games. I'm working on my own now!
This generation of AI products has been trained on the entire internet’s worth of data, which at the time of harvesting, was 100% human-generated. As more and more AI-generated content appears online, keeping those models up-to-date, or creating new ones, will become more difficult. Training models on AI-generated data leads to the AI equivalent of inbreeding. The industry has, in effect, fouled its own future larder.
"At the time of harvesting", there was already gpt2 text out there and gan-genererated images on the internet so no, it was not 100% human.
I love your focus on the current, practical uses of generative AI vs. the promised or over-hyped. Thank you.
17:47 "I don't think we need to panic, we need to hold these tech companies accountable..." SO WE DO NEED TO PANIC! Big Tech is out of control, expecting they will somehow be accountable is nuts!
15:28 as a speaker of Portuguese, I found this audio clip super interesting. It definitely sounds a bit robot-like, but the thing I found so interesting is that it sounds like someone speaking Portuguese with a fairly heavy American accent. It makes sense, since some phenomes aren't shared between the languages, but still very striking.
It's almost as if Matt spent a few days learning pronunciation basics of Portuguese, had Chat-GPT generate a transcript in Portuguese, and then read the transcript (in a robot voice) :P