Previous video on the negative side of the AI: th-cam.com/video/vQChW_jgMMM/w-d-xo.htmlsi=oFrosTUKA4saTpEA Also correction on the number of amputees, it should be 50 million not 500 million. The 500 million figure included smaller amputees like fingers and toes.
Great to see a deep dive. That's how you get to the truth and cut through the hype. The grift of people like Kurzweil is to talk at high-level generalities, glossing over nuance and equivocating. I've been critical in the past when you've made errors, such as "AI gives birth to AI" and when you fell for the FSD hype. But this video is very good.
I really appreciate you making this video, the bulk of your viewership seems to HATE ai and aren't even willing to digest what you have to say but it's only getting better and they will adapt in time.
@@johnrperry5897 It was obviously a reference to This Morning. This is an example of something called 'humour' - I'm sorry to hear that you struggle with this, but I wish you all the best in your struggles to understand normal human discourse.
@@サンゴ礁ScleractinianJohnrperry sounded like he was complimenting you or asking a question so why were you hostile to him? I didn't understand your comment either. And are people not allowed to ask? Or if he was joking, you said you liked joking so what was the problem?
big if, that. plus those corporate profits will be going down as the world loses 18% of global GDP by 2050 due to climate change disaster and shortages.
The dillema is that corporate greed is what encourages people to fund R&D in the first place. Imagine if we set a rule that cancer drugs cannot be for-profit. No one would be willing to spend billions of R&D for what's essentially a near-guarentee failure drug creation success rate.
Yeah, but even rich people understand that the current economic model is not sustainable. And they also understand that when there is too much poverty and people can no longer afford basic needs such as food, shelter, and medicine, then heads start rolling. Nobody, including rich people, want to live in a violent, unstable, and militarized world. So a great deal of effort is being put into to trying to solve this, even if it might not seem like it and it's taking a very long time. Yeah, some rich people are just flat out evil or stupid, but many are not, and they understand that corporations can't continue to behave like bacteria in a petri dish, because they will eventually run out of food/profit and then they'll all die. The corporation might not understand that as it's more like a hungry animal, but the people that make up the corporation do understand. Hence why we have things the upcoming UN Summit of the Future declaring a global crisis and stuff. Will anything significant come of it? Probably not because there are way too many petty and self absorbed people in the world and the issues are extremely complex, but the point is that if they don't change willingly then forces of nature (which includes war) will do it for them.
@@theceohq Translation: it allows for profit companies to post on it. as an internet socialist, @Planetside223 is working under the assumption for profit inherently = evil. If they are educated, they would likely have to get through the Hegelian dialectic and the entire labor theory of value to explain why, so instead, they will just argue as if the LTV is self evident and use put downs if you ask for clarification.
The chart at @5:45 is misleading in that the points for Hopper and Blackwell are FP8 and FP4. To make the comparison more fair everything should be in FP16.
@@jamesowens7176 doesn;t matter. (if) the aim was to show relative performance, than wouldn;t it have been better to shown all at the same fp count? An example i gave else where is saying a 3rd grader does better in test than a 2nd grader, but the 3rd grader took a 1st grader test vs the 2nd grader's 2nd grade test :/
That NVIDIA progress chart is extremely miss leading. Notice how it went from comparing FP16 to FP8 then FP4 scores. FP4 is significantly less resource hungry than FP16. (16 bit float point numbers vs 4 bit float point numbers. A lot less compute and precision for 4 bit floats.)
I'm glad to see that someone else caught that, it's a prime example of how to lie through numbers. All the average person sees is 'line go up'. "There are lies, damned lies and statistics"
The thing is, GPUs didn't have dedicated hardware for fp4 precision. Just like before, GTX 10 series GPUs only had fp32 and not fp16 support, but new GPUs now do. Even if I give you a model in fp16 and fp4 precision, your average consumer GPUs now will still run faster in fp16, thanks to the hardware support (provieded that VRAM is enough). So yes, even though the label is different, the practical improvement is still real.
Well how do I educate myself on this? I've been falling AI/Nvidia for several moons and your comment is the first I've seen clarify and expose this branch of AI. How do I learn this?
we are born too late for earth exploration born too early explore the universe .. we just born in the right time for the age of information ... and before we rest for good( if we are lucky or unlucky ) we can get a glimpse cyberpunk dystopia .. or really unlucky they extend your life for slavery
@@openperspective5423 doubt they would use humans for slavery. it just makes no sense. more likely we will all be uploaded on a chip to explore space while in a full dive VR simulation of anything we want.
"If corporate greed stays out of it ..." Not even a remote chance of that happening. AI works 24/7 for free. Corporations are already implementing it wherever possible. Lots of people are dismissive when someone says that AI will kill jobs. Well, it's happening right now and it's only going to intensify as AI becomes more powerful and versatile. The potential of AI is truly incredible, especially in the sciences, but the bottom line is that humans are greedy. Aside from the wealthy, who's going to be able to afford bionic limbs? Corporations will fuck everything up. When I was a kid, I was crazy excited about the future. Today? I'm content to be old. When I think about the negative potential of AI, I have no desire to find out what the future holds for humanity. Dystopia seems more likely every day.
AI does not work for free. See this excerpt from Stanford HAI Research's Artificial Intelligence Index Report for 2024: "Model training costs, as first reported in last year’s AI Index report, also continued climbing. New estimates suggest that certain frontier systems, like OpenAI’s GPT-4, cost $78 million to train. Google Gemini’s price tag came in at $191 million. By comparison, some state-of-the-art models released half a decade or so ago, namely the original transformer model (2017) and RoBERTa Large (2019), respectively cost around $900 and $160,000 to train." It should also be remembered that this AI model then has to be hosted on widely accessible hardware for consumers and businesses to use at scale, which comes with massive infrastructural and human costs. Data centers are being built en masse and municipal power grids are being pushed to produce more and more power to keep these models running. If AI worked for free, we would be seeing a LOT more of it at far cheaper prices.
I was a pharmacist a lot time and have gone into biomedical engineering; and despite how many examples you've covered, you've barely scratched the surface of how exciting some of these implications are - my knowledge too just scratches the surface But the beneits in the drug discovery and radiology part are incredible honestly We're not just talking more medications and better accuracies BUt we're talking DRAMATIC reductions in the time to do so; normal time from design to market for a drug is 10-15 years; AI has shown this could be cut down to >2 potentially (like it speeds up the early phases that much which are the very expensive parts too) ANd radiology just across the board; we're talking diseases we couldnt accurately do before like endometriosis that currently takes on average about 10 years for patients to get diagnosed I'm happy to see people with reach like you talking about how exciting the biomedical research side is I'm doing my thesis in tissue engineering (we're trying to produce artificial cancer cells for use in radiation therapy models; pretty sure we're using an LLM for the targetted radiation approach too) at the moment and it's dramnatically sped up research and design; there's a concept tissue scaffolds often rely on called auxeticity or a negative poison ratio - this is a result of how the shapes are combined and what the shapes are for a material AI models have allowed us to optimise and rapidly design these scaffolds for testing with custom models and so much more Its an exciting time to be doing cancer research
Love the optimism. I believe humanity is good. AI will help us do what we do. You're creating synthetic cancer cells? Like cells that don't want to die. I'm gonna be passed if you create the zombie apocalypse 😅 jokes
@@piranhaofserengheti4878 You do realize that the scientists will do Quality Control and Assurance right? You have any idea how, especially in developed countries they take seriously the quality if their products? Personnel with 15 to 20 long year careers in Quality departments have being FIRED and BLACKLISTED for stuff like that and you think they will just sell AI medical stuff willy nilly?
Unfortunately, "AI" is a now buzzword, a broad umbrella term for all kinds of machine learning techniques. A major technique, are neural networks, esp. in the prior "deep learning" revolution (prior to ChatGPT). Discussions about it, can be confused, esp. about energy use, dangers, ... since "AI" can mean many different things. Nowadays, they tend to be confused with LLM's. What the basic deep learning techniques offer, are pattern recognition capabilities. These have wide applicability to all kinds of scientific applications, specific niches, tools that are focussed on doing one thing well. (It's not general AI.) The specialized tools are quietly making a contribution, in the background.
I tend to disagree with the view going around lately, that LLMs somehow aren't proper "AI". They indeed are, or at least might be an important part of achieving higher level AI, like AGI, if used right. The more advances I'm seeing in AI Agents, with self reflection built in, the more I'm certain we're getting closer.
@@NikoKun LLM's is "AI". That's mostly what is meant by "AI" today, esp. given the hype as the road towards "AGI". People are thinking mainly about ChatGPT and the raft of GenAI tools. That's precisely the problem. But the field is much wider (with many other techniques).
The problem when discussing the risk/reward of AI is that it's such a dramatic shift. You can analyze it's impact as much as you want but the future will not judge the success of AI on the risks and benefits we see now, but on the risks and benefits we've not even conceived of yet.
The use of AI for the benefit of mankind in the fields of science, medicine, nature is still in it's infancy but can not be stated enough. The people that dislike AI specifically because of entertainment sector such as art, music is SO SHORTSIGHTED and actually selfish that I just ignore their general complaints....and I'm an artist myself. But I'm also a student of science and technology. AI MUST grow for a better future for humans and life in general.
How very upbeat. Id encourage you to reference every invention in the history of mankind that was supposed to make life more better. Then see how that played out. It's not that everything sucks as much as the capitalists come in and make things worse for profits. If you can answer why lightbulbs go out so fast you'll understand that the potential for abuse in any industry and with any tech is real and will translate to AI. But sure. Enjoy your AI essay writer/coder.
Have a look at what Oracle announced about their surveillance, monitoring and compliance/enforcement programs. Living in nazi occupied territory or soviet union sounds like pleasant holidays
@@davidd6660 Knowing how humankind has been over the years, I can easily imagine some countries, or terrorists using AI to develop viruses, new deadly weapons, and other ways to cause harm.
I'm at a major bank. I've seen employees post passwords and customer data into the prompt. And we know they're retaining queries and results due to federal CSAM investigations. Meaning companies like OpenAI have plaintext files with your personal information, and your bank's passwords, and it's only a matter of time until your data is stolen. It's when, not if.
there are no problems, at all, you just have to make a choice, you either use ai and llm's, or have at least "some" privacy. Not both, never, ever, in any manner. But to be honest, for most of the people - there's almost no valuable data that we didn't post by our own trusty hands in the past. No llm or facebook data steal involved. TLDR when we talking privacy in the old categories , old common sense - that train gone for good a while ago.
There is now GPT enterprise and Microsoft Copilot, which are run with contracts, thus having the liability coverage. If Microsoft and Open AI does not want big lawsuits with these kind of contracts, they better make sure the data uploaded to those models are safe. Companies already share tormendus amount of data with their suppliers and clients. All is done with the trust that the other 3rd party will work by their contract. There is literally nothing else even today that protects anyone's data but these contracts between parties. Its nothing new imo. If your company has no internal policy and training and reprimanding to colleagues about AI and its potential risks, you are not a reliable company. I agree its a privacy issue, but to me its nothing new. You think people working at companies who dont speak the local language but need to work in India as part of an SSC didnt use already google to translate some documents for their work, essentially uploading contract and other data to google? Please.... dont be delulu. This has been happening longer than you think.
Little expansion on the chip foundry section. Structures aren't really 2nm or lower in size as that can cause issues with quantum behavior of electrons at that scale, which is why the development of more efficient structures is portrayed in the media as "on the edge of physics" and the numbers are cranked down for marketing reasons. Usually the developments boil down to developing more densely packed designs or slightly better lithographic processes that are resulting in an efficiency jump (usually measured in transistor density or power consumption/compute power) that otherwise only would be achieved by downscaling classic ~40nm structure processes to e.g. 14nm, 12nm, 8nm or 2nm. Now the new software by Nvidia is used to compensate for the diffraction of UV light that is sent through a mask and "carves out" the circuits from the silicon. The mask essentially is a (partial) blueprint for all the circuits to be etched into the silicon wafer. But at these small scales the photons of the UV light are diffracted a lot by the masks they're sent through which causes blurry edges and in the worst case results in separate circuits overlapping, circuits with gaps or other faulty areas in a chip. To make these masks as precise as possible so one can increase transistor density and reduce faults, software simulates how to achieve the best design and also generates the masks for these designs, with those simulations traditionally taking up thousands of CPU hours (the mentioned 2 weeks). The AI by Nvidia is used to improve the mask software's prediction as to how the UV light will pass through the mask and increases the precision of the lithographic process while also being a lot faster (about 8h). This allows for a much quicker change of a process as you don't have to wait weeks for new masks. ASML, as the world's leading lithographic machine produce develops the machines which TSMC is using to produce Nvidia's GPU designs. TSMC uses Nvidia's new software on existing, modified machines, so it can only speed up the design process for the machines they currently have available. ASML joining in is crucial for Synopsis and TSMC to in future have machines that are both on the edge of modern engineering and capable of using Nvidia's software, else if they don't have machines that are precise enough supplied by ASML, TSMC and Synopsis will only get so far with Nvidia's AI based software optimization before running into limits again.
The largest issue with AI that hasn't gotten the proper recognition yet is the sheer amount of data it catalogs and stores whenever you interact with it. A lot of companies do not secure that data properly and allow many models to feed on that same pool of data. Security has been a core afterthought for almost all these companies because they always feel it slows them down. Not until something like one of these Medical AI apps gets compromised or starts spewing out other people's records will people really start to care.
@@NeroDefogger You are beyond delusional if you are judging it based on its current application. Look beyond the now and into decade or more into the future. It's already advancing too quickly, the only thing holding it back now is the lack of practical robots, but it will be solved. But go on, keep living in the present like a goldfish while the world is being stolen from us.
The idea of AI writing up patient notes is terrifying. Being understood and diagnosed had already gotten extremely difficult. And now doctors can check out, and let an AI that can't reason and doesn't know what it's talking about put life or death information into a patients record. And AIs tend to be biased toward maximizing profit, which will makes things even worse for patients.
Doctors are also biased towards maximizing profit, alongside their other human personal/racial/gendered/socioeconomic/etc biases. I think there's potential for some good to come of this, as evidenced by how difficult it is to be understood and diagnosed these days due to those biases.
@@Andy-fs5hv Yup, I know that from personal experience! But AIs specialize in sounding plausible, when they actually can be inaccurate to the point of spewing nonsense. Voice recognizers and expert systems have been around for decades - it's tech that's proven and adoptable. But businesses don't like them because they require smarts to set up, and maintenance once they're running. AI are a craze because they appeal to investors and require less input. They don't actually work, but businesses don't care about that.
Google is using AI to process their satellite images to precisely locate difficult to find species? Well, I’m sure they’ll be using that technology responsibly.
The problem isn’t AI. The problem is the strictly capital incentive that it’s driven by. It may be saving the kelp and helping people regain their mobility but once it’s fully shaped it’s gonna cost an arm and a leg, and more than a whole kelp farm.
Back in the early 90s, I investigated early AI and how it was being developed. I found there were two different types that were radically different. One was Neural Networks, which devised electronics to replicate a human brain. It was difficult and time-consuming, so it was basically no better than a fly's brain. The other was Expert Systems, which replicated it using software. This, of course, was simpler and easily manageable. The problem was that it wasn't very expandable or replicated the thought process very well. Years later, someone told me that the current AI tech we were using (about the time GPT came out) was the Neural Network version amplified with our new technology. Now, I'm having serious doubts about the reality of that. If it was truly Neural Network hardware designs, then current AI would be at the level of actual brains, many times over. Obviously, hardware is involved in AI to some extent, but I think it's more like a cross between NN & Expert, just boosted with better hardware. I feel that we may never actually get a good quality AI going because we didn't stick with the true Neural Networks design. Our brains don't work similarly to computer systems, so nVidia can't really be making real Neural Networks designed graphics chips. Of course, I could be wrong.
Harrison Ford: It wasn't me, it was the one-armed man! One-armed Man: It wasn't me, it was my prosthetic limb! Limb: It wasn't me, ChatGPT made me do it! ChatGPT: *I'LL DO IT AGAIN!*
@@Thatfuckingduck No, it took billions of dollars of investment and scaling up to do it. The scientists developing the tech got very wealthy doing it and have access to endless VCs who are begging them to accept their money. Greed runs the world and it's awesome, we need more of it.
@OccultDemonCassette It wont matter because economics is about to be redefined. Once enough robots are around making and fixing stuff nobody will use money any longer.
I've seen a lot of AI projects in med/biopharma that focus on narrow tasks like binding energy predictions. But these models often fail to translate into real-world results. It's like they're just regurgitating existing knowledge from the literature, without adding anything new. For instance, there was a project where a research grp tried to use AI to predict drug side effects. They fed the model tons of "journal" data, but it couldn't accurately predict (or "suggest") any side effects that hadn't already been reported. It was basically telling the story already in the data -- it was a retrieval-and-assignment aka 'inference' operation NOT prediction from learnt patterns. We need AI that can go beyond the limitations of "existing" data and provide truly innovative insights. Otherwise, it's just a shiny new tech fad that doesn't make a real difference in patient care.
It would be great to see healthcare costs decrease if advanced AI proves to be highly effective. However, my concern is that it might become another tool for corporations to cut costs and maximize profits.
anybody ever talk about health prevention? living a healthier life? creating a healthier ecosystem, lifestyle and harmonious economy??? isn't that obviously health care ?
Idk if you know, but there's already lawsuits against uber for using it for Wage Discrimination, and the same time. This new "charging algorithm" was implemented was the first time uber has everybeen profitable ever.... it's all downhill from here
As a graphic designer that's terrified of AI its hard to get over the fact this is going to dramatically decrease the amount of us required but it's hard not to appreciate that the end result of it is going to be a net benefit to us all in the long run, especially with what it's already doing for those benefiting from it already in the medical world.
@@Jplummer-y3j Something like that already exists. But that wouldn't be a good idea to attach to your body. What good is a "superhuman" power arm if the rest of you is brittle human. You will just injure yourself more except for certain minimal circumstances.
Dagogo I LOVE your content, I think I literally saw all your videos to date but I am not sure if this is me or there are others out there who feel your titles are more and more click bait-ish with no clear conclusion. I just watched this video and ended up thinking/feeling what did I even learn from this? I know it is hard to make content and find good subjects but just squeezing the AI trend does not make your content better than it used to. Your iPhone story video for e.g. is one if not the best video about it's history in my book.
I stopped watching Cold Fusion videos about 7 years ago. It's click-bait, empty platitudes, toxic positivity, low-information, poorly researched/represented, and ultimately hollow. But people will opine that "they like the music and it makes them feel good." 🤣
@@fluxonite Coldfusion does have some helpful information. I appreciate some of his positivity and different perspective instead of just fearmongering and not allowing us to have new technology. You said you quit watching but you recently came in and commented here.
This channel has grown so much, really proud of how far you got with such great content, and now great to see it evolving with real world in person portions
The thought of the tech sending "sensations" back is so scary. It's usually one-way communication, right? Bi-directional is just a scary thought to me. Wonder what those Nueralink patients are thinking (pun intended) about that?^^ cheers
I think there maybe an easy but limited way. A pressure device that is worn, back of hand, neck or back. When the device moves or make contact. it would create that sensation on a different part of the body, Including the degree of pressure, and in three axis. Pin points of light or sound waves could be translated to pressure points. No extra surgery would not be needed, but lots if learning would be.
Okay, but like, that thumbnail has changed like three times, lol! First one was the bionic arm woman. I forget the secon one but now it's an image of what I'm calling plastic kelp 😅. Admitttedly, I do think your first thumbnail gave away the surprise right from the start, so I understand wanting to make it less amniguous, but as a non content creator, it is interesting to see when a creator is trying to hit the nail on the head with a video 😅
A.I. What is the number 1 problem that plagues the world? To many people. Here is a protein that unfolds into a virus. It is carried thou vapors of petroleum,
@@DanCreaMundos Why do you think this. Yes movies like terminator resemble real AI takeover scenarios about as well as zombie movies resemble covid (IE not much). But ruling out all AI takeover scenarios in general "because movies" is also stupid. Movies aren't evidence one way or the other.
"Mr. Armchair Skeptic out there might call BS." Come on, man. Skeptics might be wrong a lot of the time when they call BS, but to patronize someone because they're skeptical (of for-profit technology, btw) is pretty low.
I was working on a video about this topic, Sarah Delagarde specifically, but when I asked her and her team what improvement AI had made I got no answers. Still no answers 9 months later. So I now believe that they have fallen for the hype of AI, raising $150,000 euros to pay for limbs that are no more capable of learning than standard prosthetics. Convince me otherwise. I may still publish my video
Just publish the video. You also may not have been given an answer because they'd rather keep you and other journos in the dark about what they are really up to. Maybe you could expose something
Almost 20 years in tech and still didn’t feel satisfied. Made a 180 into welding and am having fun with it so far. Pay isn’t as high but I’ll always have skills that AI just can’t do (at least for now) and are in demand. With that said…anyone need an IT Welder? 😅
It might get erased - but you are among the best content producers Dagogo, your contents provoke thoughts and new ideas. That is also how what you call "other side of the coin" works, it is fast-tracking a lot of things.
@@raydespoir I mean, while a funny visual, it's not a serious consideration. You never seem to worry, what happens if that automatic door closes on me, or what if I get hit my a meteorite when I'm outside? This is no different, it does nothing other than the job it's programmed for, it had no advanced reasoning other than learning motion of an arm and fingers translation. That's it. You're watching too many Hollywood movies heh, machine learning today has very very little of that aspect in it. It's always good to be vigilant, but no need for hyperbolic imagination other thab for comedy haha
@@kgmindustries8121 the one of the lady holding a knife was a little scary. Not that her arm would attack her, just that any misunderstanding or malfunction could lead her to stabbing or cutting herself, a real possiblity for real arms, and definitely possible with an AI that thinks you want to move your arm upwards while holding a big kitchen knife.
Great content as always, and as we've learned to rxpect from you, Dagogo. But it sure serms likf you're getting aggressive with all the frequent and unskippablr ads. Maybr comsider dialing the ads back, or at the very least makr thrm skippable. I can only speak for myself, but I tend to avoid videos and channels that are over the top with the ads.
In FPS multilayer games, there are many developers/companies that are working on AI team mates who you can give orders by simply saying it via your mic or typing in chat. I'll give an example -- Let's say that you are playing in a multilayer MilSim game, like 'Hell Let Loose' and you are charging the enemy position and you need help, you could ask the AI team mate "Go left and flank the enemy that are in the building ahead! Throw a grenade in the entry point before entering" I could go on, lol. It's going to be a fascinating future with AI.
Have you ever played perfect dark, N64? The enemy sims in the multiplayer mode interact and operate just like an extremely well trained unit, that almost never makes a wrong decision even against people who are by nature unpredictable, and they can react so much faster than human decision making that it's just ludicrous. And that was just one aspect of the AI implementation in a more than twenty years old Nintendo game. The reality of what we are seeing and hearing now is the result of 50 years of countless dollars and brilliant minds working intensively. It's not artificial. It's just the results of our collective intelligence coming to fruition. The world is moving faster every day and watching TH-cam only will show you the past. The current tech and future of it is moving so quickly and so shrouded in corporate and government secrecy that only those operators involved know and they are highly compartmentalizing things to ensure their success without leaks. Things to think about. I could go on for hours but surely you can extrapolate from here.
To be precise here, the "more than 550 million amputees worldwide" comment is from a Front Public Health paper last year, that includes any traumatic amputation cases. This means that if you lost a finger, you're included in the statistic, which is why the number is so mind-numbingly high. Although to be fair, the amount of major limb loss, the kind that people probably think of when they heard this number in the video to begin with, people who have lost an entire arm or leg (or sadly both) is still between 40 - 60 million cases.
@ColdFusion AI definitely has good stuff going on. What I want to know is how good the people in control of it are. Any chance you could do a vid on how wealthy the people/families in this vid, that are benefiting are?
The problem is that AI gives an excuse to lay off employees, and it makes a lot of people redundant (or will soon). There will be zero point of going to university for a lot of degrees because there will be a near impossibility for students to find work. So you have a massive collapse in the sphere of academia. Everyone can’t march I to the trades because that is directly related to supply and demand. Everyone can’t march to other majors because the problem will still remain. AI is going to eff up economies and cause massive depressions on the scale we have never seen before.
I'm personally heavily invested in that robotic future. Yes it'll disrupt employment as we understand it today... But I prefer cars to beating some horse with a whip when visiting family. Not all loss of employment is bad... Even if a lot of those horses ended up as dog food during the transition.
This disruption is the whole point, and it's one of the reasons that despite the red herring that "AI is coming to destroy", the progress on it continues. Besides, the structure of all our modern paradigms are all bunk, and what you don't realize that one bunk system is just getting replaced with another. Thanks to the ignorance of the millions upon millions of layman out here, our institution are only so brittle that they will crumble by such rapid change in advancement because people of ignorance allow such corruption to perpetuate. So despite any perceived activism towards the abolishment of AI, people should remember that their inaction against inequity is also part of the problem.
550 million amputees ???????????!!!!!!!!!!!!!???????????? that can NOT be right !!!! : 8 billion people, 550 million amputees: 1 person on 16 ???? ABSOLUTELY NOT !!!!!
Thing is, in these 8 billion there a lot of people from underdeveloped countries with a lot of manual labour with close to no automation, so a lot of amputees due to sickness or accidents. Still not sure about the number, might be an overstatement
Yeah more like 55 million worldwide, though the rest of that statement rings true about the projected numbers in the US. All the obese diabetics that end up with limb amputations from shovelling high carb high fat 💩into their bodies everyday whilst watching Fox News...
Pharmaceitical companies cutting down on drug development time is only going to benefit the shareholders. It will not make drugs cheaper or actually help people.
As someone who has studied biotech I can assure you it will benefit people cause drugs can be sent out faster. Like imagine a bacterial disease as infective as COVID-19, and instead of waiting 5 years to develop drugs, you can do it in a matter of months or less
You're completely wrong drugs will be a lot cheaper because it takes less money to build it using AI which means they can sell it for cheaper cuz they have less overhead so they make more profit. And if one company doesn't another company will that's how the free market works I build it for cheaper other companies make their product cheaper
@@EspeonMistress00 I just had to pay $4,000 for one bottle of medicine because there was no generic alternative and my insurance screwed up. How is this even remotely justifiable? Drug companies will simply hide the vast savings that "AI" brings and pocket the profits. Nothing will change. They are evil incarnate. Ask all the victims of the Sackler Family.
Man, I remember back around 2018 I bought a 1080ti off of a doctor who was experimenting with deep learning. He used the GPU he sold me to scan XRays and MRIs. He explained the whole process to me. It was mind blowing, look how far we've come in just a few years :)
As a researcher myself, I strongly support the use of AI to tackle hard problems that can benefit humanity. Rehabilitation robots, robots used in extreme conditions and others could benefit heavily with AIs. I do not support the use of AI in creative jobs. AI should be used as a way to help humanity in the long run.
As a musician and painter, the term "luddite" comes to mind. AI is far and away THE most creative tool ever devised and those who try to hide behind some mythological concept of the human "more", call it a soul, call it a spirit, call it whatever you please as special pleading is special pleading no matter what the term you use to obfuscate it, those people? They'll be left behind in history's waste bin.
Good thing this kind of AI doesn't function without a consistent feed of immense cooling and electricity. Oh but don't be luddite when you be come to expensive to hang around in the lab@matthewlawton9241.
Stringent copyright laws around AI is exactly what will bring forth the corporate dominated AI, as only the megacorps will have the money to buy the million licenses for every piece used in the training data. Taxing the AI companies and using the money to enact UBI and then embracing open source is the better approach imo, than throwing sticks in the wheels of AI, creating massive barriers for entry in hopes that artists can hobble along a few more years. It's not like the royalties off of that 0.00001% training data contribution is going to net you a living anyway. The commission will be a few dollars per piece if that.
@@MidWitPride Let's be real though, AI will become corporate dominated no matter what. It's like setting up an ISP-business, the cost of entry is far too high for any small actor to compete. It costs hundreds of millions to do a training run on a frontier model. It's not like software, where a small team with a good idea can revolutionize the industry(like whatsapp), small players just don't have the compute to compete. Open source doesn't matter all that much either, as it is far from enough for people to duplicate the models(GPT2 is open source, it's not extremely complex, honestly. but the training needed is just prohibitively expensive). Ironically open source might benefit the big guys, as if for some reason a small lab got blessed by god with a fantastic idea, and trained a model that actually made novel progress, the big players could just instantly copy the architecture, and iterate on it with their far greater access to compute. The barrier of entry that is compute is so much higher than anything copyright could ever pose, that it isn't even funny. The only way AI doesn't become corporate is if the big labs are nationalized. AI under a capitalist system is guaranteed to become corporatised.
The current rush to A.I by every business imaginable reminds me of the early days of the Internet when every company had to suddenly have a website. They are all so vague about what their A.I. is that i have no idea what they mean by it.
I am no fan of AI. I am retired and have no use for it unless for better or at least quicker video and photo editing. My son is a graduate degreed computer engineer and code writer and he has no fear for his job or future. I have no fear at the moment, but I fear for my grandkids, especially with our government already out of control. The next war may not include weapons other than supercomputing AI. I can see computers/ robots taking over most everything at some point, with or without human control.
> I can see computers/ robots taking over most everything at some point, with or without human control. My best guess, all the humans die shortly after the AI takes over. (Ie without humans in control)
von wrote, _"My son is a graduate degreed computer engineer and code writer and he has no fear for his job or future."_ Then he, or you, don't understand what's unfolding in real time. I don't mean that as a disparagement to either of you, but the rapid pace at which AI continues to progress is likely incomprehensible even among the smartest segment of the population. Part of the problem here is simple hubris: humans tend to think of themselves as the pinnacle of intelligence but this simply isn't the case. We're a lot stupider than we care to admit, and I say that as someone who from my teen years to my now retirement years, has tested quite highly on IQ tests. That's not a humble-brag, it's just offered for context that I'm not exactly some ego-driven, dopamine-seeking yahoo with a baseless opinion. In my view, the biggest challenges associated with AI have less to do with the technology itself, and more to do with the people who control its implementations. 'They' will continue to do as they've been doing for millennia, and those of us who have less agency in the matter will become unwitting victims of poor *HUMAN* choices, not directly of poor AI choices. We have met the enemy, and they are us -- by which I mean: they are human. Thousands of years of written philosophy instructing us in how to delve deeply into important matters, and ultimately it's those with the greatest deficit in empathy who rise to the top. This is an indictment not only of such individuals, but also of the collective that fails to recognize this dynamic; or, if they do recognize it, who then fails to prohibit it; because they have their own narrow-minded self-interests and cannot be bothered to understand the truth of reality -- as long as they get what they're after, Rome can burn and they'll be focused on manufacturing violins for the masses.
@@RichardHarlos Few people want to destroy the world, so if the world is destroyed by AI, it will probably be by accident. Our understanding of what code leads to what AI behaviors is rather rudimentary.
@@donaldhobson8873 wrote, _"Few people want to destroy the world"_ Ignorance is no excuse. People don't care whether what they do has long-term consequences. So, when you say that few want to destroy it, that's not nuanced enough to portray the truth of the matter. Which is why _"it will probably be by accident"_ is both true, and yet incomplete. The precautionary principle states that in the face of an unknown, one ought to err on the side of caution. AI is a vast unknown, so to race down that path as if we know where it leads when we don't, and in fact we can't... is folly... hubris... ambition... none of which give one whit about the future. So, while it may be true that they don't consciously, deliberately want to destroy the world... it's also true that they don't care whether they do. And that's a problem.
The problem with AI is the huge amount of data required for its development is stolen. It's not just image generation or LLMs. All image and language recognition AI is made using stolen date. Tesle probably gets driving data same way. Google gets all their data by stealing it, there is music too, I'm sure I'm missing other fields. The companies are getting rich with this stolen data. They say it is opensource but use it in their own tech and charge people for it. Then there are cases like adobe that make people sign an agreement to use all the data made using their software which the users had to pay for in the first place. When the industrial revolution was starting people wondered what they would do with all the free time and abundance of resource with the use of machines. None of that happened, the rich just got richer. If any part of the AI required stealing data directly or indirectly then the people should be compensated fairly, or it should be FOSS and cheap, companies shouldn't be allowed to get rich off of it. People are already losing jobs. If it goes on like this then more people will be poor and they won't even be able to afford all these new AI innovations.
Years ago AI was called 'Expert System'. CPU given a database of a 'subject'. You would submit a query, it would search that database & kick out responses. Then as now it really matters who designs & maintains that AI with the data it uses. It has access to the full internet, OK how many times have you done a search and found conflicting if not out right contradictory information? OK how will that AI deal with that? Oh trusted sites, OK is that 'trusted' site bias in some way???
The algorithm for fact-checking is always the same how many "top" sources are saying something? So unless there's a lot of "top" sources saying the same misinformation, it will usually be able to figure out what's true. But even if it doesn't you can't expect anything to 100% perfect.
Datasets that current frontier models are trained on is highly refined for quality, as the quality matters A LOT in intitial training. The data is not the same as the garbled mess a web search would get you
I always like your take on these new technologies. The kelp forest thing was my favorite part, although prosthetics and new medicines are right up there too. As always, I love your presentation style. Thanks for the show! PS: Watched this on Nebula (as always) but came back here to comment.
"Al isn't coming for human jobs" Personally, I think people should stop saying this, even as a joke or with the "yet" on it. There's no value in that narrative. It WILL take a majority of jobs, possibly by the end of this decade, and we need to start preparing for that eventuality, right NOW! I lost my career path to Al last winter, and I know half a dozen other people who've also lost jobs to Al since last year! This IS coming, and the longer we delay adapting our society and economy, the more unfair it will get. but for my robust investment portfolio that I'd built over the years with my financial advisor, Abby Joseph Cohen, I'd be hungry and homeless
Abby Joseph Cohen Services has really set the standard for others to follow, we love her here in Canada 🇨🇦 as she has been really helpful and changed lots of life's
I know this FA, Abby Joseph Cohen Services but only by her reputation at Goldman Sachs; even though she's now involved in managing portfolios and providing investmnt guidance to clients. I have been trying to get in contact since I watched her interview on WSJ last month
@@VonNothiasWell her name is 'ABBY JOSEPH COHEN SERVICES'. Just research the name. You'd find necessary details to work with a correspondence to set up an appointment.
Great share! Out of curiosity, I searched Abby Joseph Cohen Services online, found her consulting page ranked at the top, and scheduled a call session. I've read many reviews about advisors, but none seem as impressive as hers.
while replacing a graphic designer with AI probably is quicker and probably does save money. this approach leads to a lowering in quality that’s hard to measure and hard to address from the top down. New ideas have a hard stop after the person has left, all you get after that is variation. which can work for a while and in a limited way but fails everywhere else.
The negative perception of AI exists because every company is slapping "Powered by AI" on literally everything. I mean, we’re one step away from finding a light bulb with AI that can "predict" when you're about to flick the switch! Next thing you know, your toaster will be "Powered by AI" and trying to give you life advice along with your toast.
They do this with everything. After Apple popularized smartphones everything became smart. But is basically meant the given equipment, tool, device can connect to the internet. Heck, we even have smart washing powder. Nowadays I see a big rise in protein. Everything has protein in it, and they act like it's the holy grail of a fit body. "Do you want to eat a chocolate bar, this one has protein in it, so yoi don't even have to workout!". Also the electrification of every vehicle. I saw electric skateboards, rollerskates, beside the more common cars, bycicles, and scooters. Everything is smart, has protein, can transport you with using electricity as fuel and of course powered by AI.
Of course it doesn't. It's like when crypto bros feel your business "needs a blockchain Bro'." 🤣 Ask yourself what you actually took away from this video besides misrepresented information and empty platitudes. 🤣
5:41 the lithography optimisation isn’t really AI though I suspect. I’d would expect it’s using mathematical optimisation algorithms, rather than neural nets or similar AI concepts, where the individual calculations in the optimisation algorithms are parallelised on the many GPU cores.
Graph showing performance jump by nvidia generation at around 5:47 is due to support for smaller data types, eg fp8, and fp4. The reason it gets so much faster is largely memory access, there is less data that needs to be accessed/moved around, so it’s fundamentally a bit of a misleading graph.
The problem has never been AI, it has always been capitalism and "corporate greed". Capitalism creates monopolies, it is in its DNA, and AI is an oportunity to create a monopoly to rule them all.
@@fluxonite Allowed by who? Because there's definately countries that have tried other systems. Many of them are either in the process of collapsing, or have already done so. Even China is becoming increasingly capitalistic in the shadows, even as stoke nationalism by promoting their fake communism.
@@Jo1162 We know that's not true because marketing is a thing, and has been provably done to the point of deception. Not to mention buying up of any would-be competition.
Thank you so so much for making this kind of video covering the parts that get glossed over all the time. It's exhausting seeing constant bad faith and exploitation and just a complete lack of acknowledgement for people working hard to try to _actually_ make life better for people with the new technology that previously we simply didn't have a means to do.
humanity is about to hit another peak with technology through A.I. but we haven’t yet seen how this is going to work out in the long run inside our current system and society
hi Dogogo, i am a follower of your channel already for a while. Thank you again for making sucvh videos. I miss one essential discussion here: All technological improvements are bound by the ethical political decisions that are in the hands of those that can control the output of such technologies. It is not the technology per se, but for what purpose it is used. Eeven making a Paracelsus analogy, the diference of cure or poison is not the substance but the dosage. On this terms, i believe that most of us are skeptical with the bigger decisions regardings those technologies, as we have seen in human history, a.k.a atomic energy.
"New medicines with less side-effects for a cheaper price" - These companies have patents and they are not beholden to the needs of the customer, only the shareholder. As someone that works in software (which is being BOMBED by AI), it is clear that most people will not profit from AI. There will be more software and it will be worse and it will not be written by humans. It will cost the same and the only people profiting are shareholders. No technology ever has actually decreased the amount of work we have to do and made life less stressful it will only squeeze more out of the workers for the profit of the capital.
Hey man I know you guys are excited about this and all, but I just want to put down a wet towel over your heads, LLMs and classical neural networks are not the same, a bit similar in how they run but NOT the same, LLMs are only useful for *certain* tasks, one of them certainly is NOT prosthetics, so mixing them up in a video is misleading at best. Why do I say that? because if you take out LLMs and what they are useful for, which is basically prediction engines that are fairly slow in execution (ideally for text where the amount of possibilities is less), prosthetics are not fit for that, you require in real time movement, for that you have to use either direct nerve connections or a sort of real time AI, ideally, a very very good near AGI AI because there are inherent risks to it. If we assume that, then, there has been no actual "lightning speed" porgress to it, it's been the same "slow and steady" AI progress, which while it is somewhat happening, sure, it's definitely not gonna come out tomorrow. LLM investing companies outside of Google have near to nothing going on regarding non LLM research and as such as actually doing 0 progress towards artificial general intelligence, which will NOT be an LLM, LLMs will NOT be AGI do not let scammers trick you, read up how they work, read the papers, use them, the most up to date models in real world scenarios, listen to the head of meta AI talk to Lex Friedman about it as well, you will quickly form a conclussion that the space is selling a bunch of vaporware that doesn't exist.
When it comes to LLMs, AGI doesn't have to be sentient, it has to be useful. And it turns out, language is a way of storing ideas, that lets them be manipulated, and text works fine as a method of output. Honestly, we are closest to reaching AGI with LLM's, even if it doesn't look like you imagine, and even if it's slow, that doesn't mean it isn't useful. For example, where it can write an entire programming project based on specs, which we are pretty close to. It's flawed at the moment, but with better models and training, there is nothing in that which couldn't be done by an LLM. Quite frankly, it doesn't matter if it actually "Understands" or if it's just a Chinese room. Who cares? If it is able to produce quality results, that's all that should matter. LLM's are able to act like they understand, and that's quite good enough. And I'm aware of how they work, it's my professional field. I have probably read more papers than you have on this. Just because someone arrives at a different conclusion doesn't mean they know less than you. (Also, he never said LLMs powered prosthetics, so I don't know where you are getting that.)
You don’t need agi to create a disruptive technology utilizing ai. Agi is a world changer, absolutely, but the world could change with a highly useful ai system regardless of its agi or sgi.
Someone only reacted to the title and didn't watch the video....none of these shown are language models. He mentioned at the start that he considered "generative A.I" to be a different topic from what this video was covering. "Google have near to nothing going on regarding non-LLM research..." Did you forget about Google Deepmind and their AlphaGo model that helped kick off the A.I hype before LLMs showed up?
Amazing video, thank you for making this and pooling the sources. Wish there were more videos on this topic. "Other side of AI" type stuff. It's annoying when I bring up AI and ppl instantly have an opinion on ChatGPT and etc... I'd say that these conversational LLMs are still pretty stupid. Literally these current LLMs don't have "logic", they base answers on weights. Literally guessing what you want hear based on the weights (yes and other parameters). I wish there were more videos that discuss AI being applied in science and tech fields and not chat bots lol.
I have Chat GPT 4o and right before watching this video, i had 2 hours of conversation with it. Compared to talking with friends, it was way more most insightful and productive. At this point, it is not like talking to a human, it is better.
@@jean-claudelemieux8877 I couldn't agree with you more, Jean Claude. Nevertheless, may I point out that you are rightly considering the word "friend" in the deepest human sense of the term? (I think that IS what sushantkarki seems to have missed with his apparently reckless, yet otherwise curious comment, wouldn't you agree?). Goes without saying: I really like your vision of true human friendship (Cheers!! :)
@@edwardenglishonline I do not believe there is something to say about true human friendship here. I was simply using the, in my opinion, common sense of what is a friend.
The "Hype" also leads to significant investments from the biggest companies in the world. The AI haters are the equivalent of the people back in the day who said computers are all a waste of money or cars will never be a viable form of transportation. I don't get it, the potential of this technology is limitless.
no, iron fist era is coming - AI main task is surveillance, censorship, compliance monitoring and enforcement. have a look at what Oracle is doing and what all agencies like WEF and govts are legislating
What a completely saccharine naive perspective. 🤣 Yes I'm sure "the best companies in the world" plan to usher in a utopia for all of humanity! "Think of all the possibilities!" 😉
I'm a Gen-X'r and 59. My background is in programming. Since the early 80's I've known that technology would overtake many jobs done by humans. Civilization is going to advance. Unfortunately there are those that refuse to live in the future.
It's sad that a primarily technical TH-cam channel has to make a video defending the premise that AI can be a good thing. So many people piled on the hate train for something that's going to hit every aspect of our lives, hard, and somehow are blind to what's just on the horizon.
yes people are so emotional and in denial. I have mostly hidden comments on youtube and even tech websites because people seem to lack any sense of how big the shift will be eventually and are just clinging to any examples they can of LLMs failing
Previous video on the negative side of the AI: th-cam.com/video/vQChW_jgMMM/w-d-xo.htmlsi=oFrosTUKA4saTpEA
Also correction on the number of amputees, it should be 50 million not 500 million. The 500 million figure included smaller amputees like fingers and toes.
Great to see a deep dive. That's how you get to the truth and cut through the hype. The grift of people like Kurzweil is to talk at high-level generalities, glossing over nuance and equivocating.
I've been critical in the past when you've made errors, such as "AI gives birth to AI" and when you fell for the FSD hype. But this video is very good.
Great content
Ehm, why does CSIRO have similar name and logo to CISCO? And why does IMAS have similar logo to ATandT? :D
@dalimillazan2877
CSIRO was founded in 1916. Cisco was founded in 1984. Maybe ask Cisco.
I really appreciate you making this video, the bulk of your viewership seems to HATE ai and aren't even willing to digest what you have to say but it's only getting better and they will adapt in time.
"If corporate greed doesn't get to it"
If my grandmother had wheels, she would have been a bike.
Classic cooking show remark lol
If this was an original comment, you’d be considered clever
@@johnrperry5897 It was obviously a reference to This Morning. This is an example of something called 'humour' - I'm sorry to hear that you struggle with this, but I wish you all the best in your struggles to understand normal human discourse.
@@サンゴ礁ScleractinianJohnrperry sounded like he was complimenting you or asking a question so why were you hostile to him?
I didn't understand your comment either.
And are people not allowed to ask?
Or if he was joking, you said you liked joking so what was the problem?
@@user-gu9yq5sj7c It's not a comment. Johnrperrt is the one being hostile.
The keywords being "if corporate greed doesn't get to it..."
big if, that.
plus those corporate profits will be going down as the world loses 18% of global GDP by 2050 due to climate change disaster and shortages.
It already did
The dillema is that corporate greed is what encourages people to fund R&D in the first place. Imagine if we set a rule that cancer drugs cannot be for-profit. No one would be willing to spend billions of R&D for what's essentially a near-guarentee failure drug creation success rate.
So, never then. Unless we have an exceptionally brutal uprising.
Yeah, but even rich people understand that the current economic model is not sustainable. And they also understand that when there is too much poverty and people can no longer afford basic needs such as food, shelter, and medicine, then heads start rolling. Nobody, including rich people, want to live in a violent, unstable, and militarized world. So a great deal of effort is being put into to trying to solve this, even if it might not seem like it and it's taking a very long time. Yeah, some rich people are just flat out evil or stupid, but many are not, and they understand that corporations can't continue to behave like bacteria in a petri dish, because they will eventually run out of food/profit and then they'll all die. The corporation might not understand that as it's more like a hungry animal, but the people that make up the corporation do understand. Hence why we have things the upcoming UN Summit of the Future declaring a global crisis and stuff. Will anything significant come of it? Probably not because there are way too many petty and self absorbed people in the world and the issues are extremely complex, but the point is that if they don't change willingly then forces of nature (which includes war) will do it for them.
The humor of a video about AI innovations and the promotion of a job searching website is not lost on me.
80.000 hours isnt a job searching website (primarily) and is hella based
80,000 hours is a billionaire political propagandist group about effective altruism, accerlationsim, and longtermism. No cap.
@@Planetside223 So how is it less of a valuable job search tool?
@@theceohq Translation: it allows for profit companies to post on it. as an internet socialist, @Planetside223 is working under the assumption for profit inherently = evil. If they are educated, they would likely have to get through the Hegelian dialectic and the entire labor theory of value to explain why, so instead, they will just argue as if the LTV is self evident and use put downs if you ask for clarification.
Yeah tech bros are particularly good at telling stories against themselves.
The chart at @5:45 is misleading in that the points for Hopper and Blackwell are FP8 and FP4. To make the comparison more fair everything should be in FP16.
This needs to be boosted
@@slax4884 If only there was a button for that...
As I understand it, FP16 isn't required for neural nets. IIRC Tesla is training autonomous driving for their cars with FP8.
@@jamesowens7176 doesn;t matter. (if) the aim was to show relative performance, than wouldn;t it have been better to shown all at the same fp count?
An example i gave else where is saying a 3rd grader does better in test than a 2nd grader, but the 3rd grader took a 1st grader test vs the 2nd grader's 2nd grade test :/
To be fair we are making better use out of lower precision in training and inference (mores now than in the past), but yeah it is misleading.
That NVIDIA progress chart is extremely miss leading. Notice how it went from comparing FP16 to FP8 then FP4 scores. FP4 is significantly less resource hungry than FP16. (16 bit float point numbers vs 4 bit float point numbers. A lot less compute and precision for 4 bit floats.)
I'm glad to see that someone else caught that, it's a prime example of how to lie through numbers. All the average person sees is 'line go up'.
"There are lies, damned lies and statistics"
The thing is, GPUs didn't have dedicated hardware for fp4 precision.
Just like before, GTX 10 series GPUs only had fp32 and not fp16 support, but new GPUs now do.
Even if I give you a model in fp16 and fp4 precision, your average consumer GPUs now will still run faster in fp16, thanks to the hardware support (provieded that VRAM is enough).
So yes, even though the label is different, the practical improvement is still real.
Well how do I educate myself on this? I've been falling AI/Nvidia for several moons and your comment is the first I've seen clarify and expose this branch of AI. How do I learn this?
@@h.m.chuang0224 so then why not just show fp16 performance?
this is still massaging of numbers :/
@@Ex14 Because it *is* improvement? Why list the driving speed of the new car instead of writing the walking speed of a human?
Finally getting that cyberpunk with our dystopia.
Nice.
we are born too late for earth exploration born too early explore the universe .. we just born in the right time for the age of information ... and before we rest for good( if we are lucky or unlucky ) we can get a glimpse cyberpunk dystopia .. or really unlucky they extend your life for slavery
@@openperspective5423the biosphere, and therefore civilisation, is on the verge of collapse. All this computer is just speeding up the process.
lol. That’s some funny 💩.
@@openperspective5423 doubt they would use humans for slavery. it just makes no sense. more likely we will all be uploaded on a chip to explore space while in a full dive VR simulation of anything we want.
"If corporate greed stays out of it ..."
Not even a remote chance of that happening. AI works 24/7 for free. Corporations are already implementing it wherever possible.
Lots of people are dismissive when someone says that AI will kill jobs. Well, it's happening right now and it's only going to intensify as AI becomes more powerful and versatile. The potential of AI is truly incredible, especially in the sciences, but the bottom line is that humans are greedy. Aside from the wealthy, who's going to be able to afford bionic limbs? Corporations will fuck everything up.
When I was a kid, I was crazy excited about the future. Today? I'm content to be old. When I think about the negative potential of AI, I have no desire to find out what the future holds for humanity. Dystopia seems more likely every day.
Look up for the work of Peter Joseph, Zeitgeist and Inter reflections small light in this darkness.
Remember:
It's not the technology, it's WHO OWNS THE TECHNOLOGY.
The super-rich and corporations are NOT your friends.
AI does not work for free. See this excerpt from Stanford HAI Research's Artificial Intelligence Index Report for 2024:
"Model training costs, as first reported in last year’s AI Index report, also continued climbing. New estimates suggest that certain frontier systems, like OpenAI’s GPT-4, cost $78 million to train. Google Gemini’s price tag came in at $191 million. By comparison, some state-of-the-art models released half a decade or so ago, namely the original transformer model (2017) and RoBERTa Large (2019), respectively cost around $900 and $160,000 to train."
It should also be remembered that this AI model then has to be hosted on widely accessible hardware for consumers and businesses to use at scale, which comes with massive infrastructural and human costs. Data centers are being built en masse and municipal power grids are being pushed to produce more and more power to keep these models running. If AI worked for free, we would be seeing a LOT more of it at far cheaper prices.
Well, when automatic elevators were invented, elevator operators lost their jobs.
Rich tech people suggesting we regulate AI may be saying that for more than altruistic reasons.
I was a pharmacist a lot time and have gone into biomedical engineering; and despite how many examples you've covered, you've barely scratched the surface of how exciting some of these implications are - my knowledge too just scratches the surface
But the beneits in the drug discovery and radiology part are incredible honestly
We're not just talking more medications and better accuracies
BUt we're talking DRAMATIC reductions in the time to do so; normal time from design to market for a drug is 10-15 years; AI has shown this could be cut down to >2 potentially (like it speeds up the early phases that much which are the very expensive parts too)
ANd radiology just across the board; we're talking diseases we couldnt accurately do before like endometriosis that currently takes on average about 10 years for patients to get diagnosed
I'm happy to see people with reach like you talking about how exciting the biomedical research side is
I'm doing my thesis in tissue engineering (we're trying to produce artificial cancer cells for use in radiation therapy models; pretty sure we're using an LLM for the targetted radiation approach too)
at the moment and it's dramnatically sped up research and design; there's a concept tissue scaffolds often rely on called auxeticity or a negative poison ratio - this is a result of how the shapes are combined and what the shapes are for a material
AI models have allowed us to optimise and rapidly design these scaffolds for testing with custom models and so much more
Its an exciting time to be doing cancer research
Love the optimism. I believe humanity is good. AI will help us do what we do.
You're creating synthetic cancer cells? Like cells that don't want to die. I'm gonna be passed if you create the zombie apocalypse 😅
jokes
Ah yes, why do we need all those years of clinical trials, it was designed by AI. It's safe and effective because LLM said so.
@@piranhaofserengheti4878 yes, because that's how scientists think and work 🤦
In terms of biomedical and just solving the issues that plague people, I'm optimistic for AI.
For my future employment though, not so much
@@piranhaofserengheti4878 You do realize that the scientists will do Quality Control and Assurance right?
You have any idea how, especially in developed countries they take seriously the quality if their products?
Personnel with 15 to 20 long year careers in Quality departments have being FIRED and BLACKLISTED for stuff like that and you think they will just sell AI medical stuff willy nilly?
Been a subsriber since the channel had 30k subs. Still my favorite channel on TH-cam. Please keep up the amazing work!
Unfortunately, "AI" is a now buzzword, a broad umbrella term for all kinds of machine learning techniques. A major technique, are neural networks, esp. in the prior "deep learning" revolution (prior to ChatGPT). Discussions about it, can be confused, esp. about energy use, dangers, ... since "AI" can mean many different things. Nowadays, they tend to be confused with LLM's. What the basic deep learning techniques offer, are pattern recognition capabilities. These have wide applicability to all kinds of scientific applications, specific niches, tools that are focussed on doing one thing well. (It's not general AI.) The specialized tools are quietly making a contribution, in the background.
It has always been a broad umbrella term, hasn't it? Machine learning is a subfield of AI and deep learning is a subfield of machine learning.
I tend to disagree with the view going around lately, that LLMs somehow aren't proper "AI". They indeed are, or at least might be an important part of achieving higher level AI, like AGI, if used right. The more advances I'm seeing in AI Agents, with self reflection built in, the more I'm certain we're getting closer.
@@NikoKun LLM's is "AI". That's mostly what is meant by "AI" today, esp. given the hype as the road towards "AGI". People are thinking mainly about ChatGPT and the raft of GenAI tools. That's precisely the problem. But the field is much wider (with many other techniques).
An LLM is literally a neural network
bait, son... OP is bait...
The problem when discussing the risk/reward of AI is that it's such a dramatic shift. You can analyze it's impact as much as you want but the future will not judge the success of AI on the risks and benefits we see now, but on the risks and benefits we've not even conceived of yet.
A true black swan event
The use of AI for the benefit of mankind in the fields of science, medicine, nature is still in it's infancy but can not be stated enough. The people that dislike AI specifically because of entertainment sector such as art, music is SO SHORTSIGHTED and actually selfish that I just ignore their general complaints....and I'm an artist myself. But I'm also a student of science and technology. AI MUST grow for a better future for humans and life in general.
How very upbeat. Id encourage you to reference every invention in the history of mankind that was supposed to make life more better. Then see how that played out. It's not that everything sucks as much as the capitalists come in and make things worse for profits.
If you can answer why lightbulbs go out so fast you'll understand that the potential for abuse in any industry and with any tech is real and will translate to AI.
But sure. Enjoy your AI essay writer/coder.
@@davidd6660 That sand around your head must feel nice.
Have a look at what Oracle announced about their surveillance, monitoring and compliance/enforcement programs. Living in nazi occupied territory or soviet union sounds like pleasant holidays
@@davidd6660 Knowing how humankind has been over the years, I can easily imagine some countries, or terrorists using AI to develop viruses, new deadly weapons, and other ways to cause harm.
The problem is capitalism
Please also make a video about Hoody or similar 3rd party, there is almost no video about the massive privacy issue of LLMs
I'm at a major bank. I've seen employees post passwords and customer data into the prompt. And we know they're retaining queries and results due to federal CSAM investigations. Meaning companies like OpenAI have plaintext files with your personal information, and your bank's passwords, and it's only a matter of time until your data is stolen. It's when, not if.
there are no problems, at all, you just have to make a choice, you either use ai and llm's, or have at least "some" privacy. Not both, never, ever, in any manner. But to be honest, for most of the people - there's almost no valuable data that we didn't post by our own trusty hands in the past. No llm or facebook data steal involved. TLDR when we talking privacy in the old categories , old common sense - that train gone for good a while ago.
That's why it is mentioned in LLM that don't put personal data@@KevinJDildonik
I see several different search results when looking up Hoody. What is the exact site?
There is now GPT enterprise and Microsoft Copilot, which are run with contracts, thus having the liability coverage. If Microsoft and Open AI does not want big lawsuits with these kind of contracts, they better make sure the data uploaded to those models are safe. Companies already share tormendus amount of data with their suppliers and clients. All is done with the trust that the other 3rd party will work by their contract. There is literally nothing else even today that protects anyone's data but these contracts between parties. Its nothing new imo.
If your company has no internal policy and training and reprimanding to colleagues about AI and its potential risks, you are not a reliable company. I agree its a privacy issue, but to me its nothing new. You think people working at companies who dont speak the local language but need to work in India as part of an SSC didnt use already google to translate some documents for their work, essentially uploading contract and other data to google? Please.... dont be delulu. This has been happening longer than you think.
Little expansion on the chip foundry section.
Structures aren't really 2nm or lower in size as that can cause issues with quantum behavior of electrons at that scale, which is why the development of more efficient structures is portrayed in the media as "on the edge of physics" and the numbers are cranked down for marketing reasons. Usually the developments boil down to developing more densely packed designs or slightly better lithographic processes that are resulting in an efficiency jump (usually measured in transistor density or power consumption/compute power) that otherwise only would be achieved by downscaling classic ~40nm structure processes to e.g. 14nm, 12nm, 8nm or 2nm.
Now the new software by Nvidia is used to compensate for the diffraction of UV light that is sent through a mask and "carves out" the circuits from the silicon. The mask essentially is a (partial) blueprint for all the circuits to be etched into the silicon wafer. But at these small scales the photons of the UV light are diffracted a lot by the masks they're sent through which causes blurry edges and in the worst case results in separate circuits overlapping, circuits with gaps or other faulty areas in a chip. To make these masks as precise as possible so one can increase transistor density and reduce faults, software simulates how to achieve the best design and also generates the masks for these designs, with those simulations traditionally taking up thousands of CPU hours (the mentioned 2 weeks). The AI by Nvidia is used to improve the mask software's prediction as to how the UV light will pass through the mask and increases the precision of the lithographic process while also being a lot faster (about 8h). This allows for a much quicker change of a process as you don't have to wait weeks for new masks.
ASML, as the world's leading lithographic machine produce develops the machines which TSMC is using to produce Nvidia's GPU designs. TSMC uses Nvidia's new software on existing, modified machines, so it can only speed up the design process for the machines they currently have available. ASML joining in is crucial for Synopsis and TSMC to in future have machines that are both on the edge of modern engineering and capable of using Nvidia's software, else if they don't have machines that are precise enough supplied by ASML, TSMC and Synopsis will only get so far with Nvidia's AI based software optimization before running into limits again.
I don’t often comment but just wanted to say I absolutely love your editing and whole setup. Your videos on AI are top notch
The largest issue with AI that hasn't gotten the proper recognition yet is the sheer amount of data it catalogs and stores whenever you interact with it. A lot of companies do not secure that data properly and allow many models to feed on that same pool of data. Security has been a core afterthought for almost all these companies because they always feel it slows them down. Not until something like one of these Medical AI apps gets compromised or starts spewing out other people's records will people really start to care.
To quote one of the memes i saw recently
"I want AI to do my laundry and dishes so i can focus on work, not the other way around"
it can do laundry and dishes and can't do (at least my and most people's) work, so it is the good way around, dunno what you are talking about
@@NeroDefogger You are beyond delusional if you are judging it based on its current application. Look beyond the now and into decade or more into the future.
It's already advancing too quickly, the only thing holding it back now is the lack of practical robots, but it will be solved. But go on, keep living in the present like a goldfish while the world is being stolen from us.
@@NeroDefogger wtf?
nahhh fk this, i dont want to work wtf? ai can take every job i dont care
@@wayfa13 what?
The idea of AI writing up patient notes is terrifying. Being understood and diagnosed had already gotten extremely difficult. And now doctors can check out, and let an AI that can't reason and doesn't know what it's talking about put life or death information into a patients record. And AIs tend to be biased toward maximizing profit, which will makes things even worse for patients.
Doctors are also biased towards maximizing profit, alongside their other human personal/racial/gendered/socioeconomic/etc biases. I think there's potential for some good to come of this, as evidenced by how difficult it is to be understood and diagnosed these days due to those biases.
Human doctors make mistakes - a lot of
@@Andy-fs5hv Yup, I know that from personal experience! But AIs specialize in sounding plausible, when they actually can be inaccurate to the point of spewing nonsense.
Voice recognizers and expert systems have been around for decades - it's tech that's proven and adoptable. But businesses don't like them because they require smarts to set up, and maintenance once they're running.
AI are a craze because they appeal to investors and require less input. They don't actually work, but businesses don't care about that.
@@darkwing3713 I use chatGPT and it’s incredibly effective I must say
@Andy-fs5hv Yes, but human doctors can be sued for their mistakes. AIs can't.
whatsinmy AI fixes this. "Are we all wrong about AI?"
Google is using AI to process their satellite images to precisely locate difficult to find species? Well, I’m sure they’ll be using that technology responsibly.
The military contractor wing of Google won't be using this tech for anything bad.
@@OccultDemonCassette Well said! Of course not!
/end sarcasm
The problem isn’t AI. The problem is the strictly capital incentive that it’s driven by. It may be saving the kelp and helping people regain their mobility but once it’s fully shaped it’s gonna cost an arm and a leg, and more than a whole kelp farm.
Back in the early 90s, I investigated early AI and how it was being developed. I found there were two different types that were radically different. One was Neural Networks, which devised electronics to replicate a human brain. It was difficult and time-consuming, so it was basically no better than a fly's brain. The other was Expert Systems, which replicated it using software. This, of course, was simpler and easily manageable. The problem was that it wasn't very expandable or replicated the thought process very well.
Years later, someone told me that the current AI tech we were using (about the time GPT came out) was the Neural Network version amplified with our new technology. Now, I'm having serious doubts about the reality of that. If it was truly Neural Network hardware designs, then current AI would be at the level of actual brains, many times over. Obviously, hardware is involved in AI to some extent, but I think it's more like a cross between NN & Expert, just boosted with better hardware. I feel that we may never actually get a good quality AI going because we didn't stick with the true Neural Networks design.
Our brains don't work similarly to computer systems, so nVidia can't really be making real Neural Networks designed graphics chips.
Of course, I could be wrong.
Harrison Ford: It wasn't me, it was the one-armed man!
One-armed Man: It wasn't me, it was my prosthetic limb!
Limb: It wasn't me, ChatGPT made me do it!
ChatGPT: *I'LL DO IT AGAIN!*
😂
Brooooooo
" If corporate greed can stay out of it"....that's not going to happen. GREED rules the world.
it created all this tech
@@tuckerbugeater Actually they were created by the the scientists and engineers.
@@Thatfuckingduck No, it took billions of dollars of investment and scaling up to do it. The scientists developing the tech got very wealthy doing it and have access to endless VCs who are begging them to accept their money. Greed runs the world and it's awesome, we need more of it.
@@CarpenterBrothersociopathic nightmare person.
@OccultDemonCassette It wont matter because economics is about to be redefined. Once enough robots are around making and fixing stuff nobody will use money any longer.
I've seen a lot of AI projects in med/biopharma that focus on narrow tasks like binding energy predictions. But these models often fail to translate into real-world results. It's like they're just regurgitating existing knowledge from the literature, without adding anything new. For instance, there was a project where a research grp tried to use AI to predict drug side effects. They fed the model tons of "journal" data, but it couldn't accurately predict (or "suggest") any side effects that hadn't already been reported. It was basically telling the story already in the data -- it was a retrieval-and-assignment aka 'inference' operation NOT prediction from learnt patterns. We need AI that can go beyond the limitations of "existing" data and provide truly innovative insights. Otherwise, it's just a shiny new tech fad that doesn't make a real difference in patient care.
Man I'm so glad there's people solving all these problems in the world ☮. If it was left up to me, we'd be cooked.
Don't sell yourself short, Thanos.
“ if corporate greed stays out of it” hahahaa this is going to be fucked up too. Just give them time
It would be great to see healthcare costs decrease if advanced AI proves to be highly effective.
However, my concern is that it might become another tool for corporations to cut costs and maximize profits.
Maybe keep an option open to travel somewhere where competition is permitted or encouraged.
AI doesnt do anything for healthcare though, at best it'll organize the info better but that's only if it's reliable enough for the task
anybody ever talk about health prevention? living a healthier life? creating a healthier ecosystem, lifestyle and harmonious economy??? isn't that obviously health care ?
Idk if you know, but there's already lawsuits against uber for using it for Wage Discrimination, and the same time. This new "charging algorithm" was implemented was the first time uber has everybeen profitable ever.... it's all downhill from here
As a graphic designer that's terrified of AI its hard to get over the fact this is going to dramatically decrease the amount of us required but it's hard not to appreciate that the end result of it is going to be a net benefit to us all in the long run, especially with what it's already doing for those benefiting from it already in the medical world.
AI may create new job opportunities, for example creating training data in a specific style
I can't wait to get the bionic arm with superhuman strength
Well if you want one, me and my brother are about to go into the basement draw transmutation circle and try to get back our mom 🤷♂️
You'd get an exoskeleton rather than a new arm.
Just make sure you remove your current arms safely.
@@edd2184 I think you'd probably regret doing that...
@@Jplummer-y3j Something like that already exists. But that wouldn't be a good idea to attach to your body. What good is a "superhuman" power arm if the rest of you is brittle human. You will just injure yourself more except for certain minimal circumstances.
Really glad you're covering this right now! I've been wondering about the use-cases we're not hearing about
Dagogo I LOVE your content, I think I literally saw all your videos to date but I am not sure if this is me or there are others out there who feel your titles are more and more click bait-ish with no clear conclusion. I just watched this video and ended up thinking/feeling what did I even learn from this?
I know it is hard to make content and find good subjects but just squeezing the AI trend does not make your content better than it used to. Your iPhone story video for e.g. is one if not the best video about it's history in my book.
I stopped watching Cold Fusion videos about 7 years ago. It's click-bait, empty platitudes, toxic positivity, low-information, poorly researched/represented, and ultimately hollow. But people will opine that "they like the music and it makes them feel good." 🤣
@@fluxoniteyour commenting this on a Cold Fusion video brother 😂😂
@@fluxonite Coldfusion does have some helpful information. I appreciate some of his positivity and different perspective instead of just fearmongering and not allowing us to have new technology.
You said you quit watching but you recently came in and commented here.
This channel has grown so much, really proud of how far you got with such great content, and now great to see it evolving with real world in person portions
The thought of the tech sending "sensations" back is so scary.
It's usually one-way communication, right? Bi-directional is just a scary thought to me.
Wonder what those Nueralink patients are thinking (pun intended) about that?^^
cheers
I think there maybe an easy but limited way. A pressure device that is worn, back of hand, neck or back.
When the device moves or make contact. it would create that sensation on a different part of the body,
Including the degree of pressure, and in three axis.
Pin points of light or sound waves could be translated to pressure points.
No extra surgery would not be needed, but lots if learning would be.
Okay, but like, that thumbnail has changed like three times, lol! First one was the bionic arm woman. I forget the secon one but now it's an image of what I'm calling plastic kelp 😅. Admitttedly, I do think your first thumbnail gave away the surprise right from the start, so I understand wanting to make it less amniguous, but as a non content creator, it is interesting to see when a creator is trying to hit the nail on the head with a video 😅
I just hope AI can help us faster than we can destroy ourselves
A.I. What is the number 1 problem that plagues the world?
To many people. Here is a protein that unfolds into a virus. It is carried thou vapors of petroleum,
AI can destroy us faster than we can respond, if it's smart.
@@donaldhobson8873 you've seen too many movies, anything that happens will be exclusively caused by humans
@@DanCreaMundos
Why do you think this.
Yes movies like terminator resemble real AI takeover scenarios about as well as zombie movies resemble covid (IE not much).
But ruling out all AI takeover scenarios in general "because movies" is also stupid. Movies aren't evidence one way or the other.
"Mr. Armchair Skeptic out there might call BS." Come on, man. Skeptics might be wrong a lot of the time when they call BS, but to patronize someone because they're skeptical (of for-profit technology, btw) is pretty low.
I was working on a video about this topic, Sarah Delagarde specifically, but when I asked her and her team what improvement AI had made I got no answers. Still no answers 9 months later. So I now believe that they have fallen for the hype of AI, raising $150,000 euros to pay for limbs that are no more capable of learning than standard prosthetics. Convince me otherwise. I may still publish my video
Very interesting
Just publish the video. You also may not have been given an answer because they'd rather keep you and other journos in the dark about what they are really up to. Maybe you could expose something
That's exactly why everything is AI now, to pump up the stock and get free publicity from AI-written news articles. Basically AI is promoting AI.
Almost 20 years in tech and still didn’t feel satisfied. Made a 180 into welding and am having fun with it so far. Pay isn’t as high but I’ll always have skills that AI just can’t do (at least for now) and are in demand. With that said…anyone need an IT Welder? 😅
Skynet needs you to build a T-800
Aren’t welding robots the most common type? Not sure you’ve thought this through 😬
@@_Agent_86 try to get a robot to weld pipe or maybe some underwater rig repairs. Definitely thought it through.
It might get erased - but you are among the best content producers Dagogo, your contents provoke thoughts and new ideas. That is also how what you call "other side of the coin" works, it is fast-tracking a lot of things.
I didn't think it was BS until you quoted Mckinsey.
Production quality is so good. Kudos 😀
That thumbnail made me think of Peter Pettigrew. Imagine getting choked out by your own arm but instead of magic it's rogue AI. Scary.
@@raydespoir I mean, while a funny visual, it's not a serious consideration. You never seem to worry, what happens if that automatic door closes on me, or what if I get hit my a meteorite when I'm outside? This is no different, it does nothing other than the job it's programmed for, it had no advanced reasoning other than learning motion of an arm and fingers translation. That's it. You're watching too many Hollywood movies heh, machine learning today has very very little of that aspect in it. It's always good to be vigilant, but no need for hyperbolic imagination other thab for comedy haha
@@kgmindustries8121 the one of the lady holding a knife was a little scary. Not that her arm would attack her, just that any misunderstanding or malfunction could lead her to stabbing or cutting herself, a real possiblity for real arms, and definitely possible with an AI that thinks you want to move your arm upwards while holding a big kitchen knife.
Great content as always, and as we've learned to rxpect from you, Dagogo. But it sure serms likf you're getting aggressive with all the frequent and unskippablr ads. Maybr comsider dialing the ads back, or at the very least makr thrm skippable.
I can only speak for myself, but I tend to avoid videos and channels that are over the top with the ads.
In FPS multilayer games, there are many developers/companies that are working on AI team mates who you can give orders by simply saying it via your mic or typing in chat.
I'll give an example -- Let's say that you are playing in a multilayer MilSim game, like 'Hell Let Loose' and you are charging the enemy position and you need help, you could ask the AI team mate "Go left and flank the enemy that are in the building ahead! Throw a grenade in the entry point before entering"
I could go on, lol. It's going to be a fascinating future with AI.
Have you ever played perfect dark, N64? The enemy sims in the multiplayer mode interact and operate just like an extremely well trained unit, that almost never makes a wrong decision even against people who are by nature unpredictable, and they can react so much faster than human decision making that it's just ludicrous. And that was just one aspect of the AI implementation in a more than twenty years old Nintendo game. The reality of what we are seeing and hearing now is the result of 50 years of countless dollars and brilliant minds working intensively. It's not artificial. It's just the results of our collective intelligence coming to fruition. The world is moving faster every day and watching TH-cam only will show you the past. The current tech and future of it is moving so quickly and so shrouded in corporate and government secrecy that only those operators involved know and they are highly compartmentalizing things to ensure their success without leaks. Things to think about. I could go on for hours but surely you can extrapolate from here.
10:50 22:45 The way you handled this section is a masterclass in skill and patience. Respect! 🎓
To be precise here, the "more than 550 million amputees worldwide" comment is from a Front Public Health paper last year, that includes any traumatic amputation cases. This means that if you lost a finger, you're included in the statistic, which is why the number is so mind-numbingly high.
Although to be fair, the amount of major limb loss, the kind that people probably think of when they heard this number in the video to begin with, people who have lost an entire arm or leg (or sadly both) is still between 40 - 60 million cases.
@ColdFusion AI definitely has good stuff going on. What I want to know is how good the people in control of it are. Any chance you could do a vid on how wealthy the people/families in this vid, that are benefiting are?
The problem is that AI gives an excuse to lay off employees, and it makes a lot of people redundant (or will soon).
There will be zero point of going to university for a lot of degrees because there will be a near impossibility for students to find work. So you have a massive collapse in the sphere of academia.
Everyone can’t march I to the trades because that is directly related to supply and demand. Everyone can’t march to other majors because the problem will still remain.
AI is going to eff up economies and cause massive depressions on the scale we have never seen before.
What makes you think there will not be more Ai research jobs created?
@@jeevan88888because AI will be doing that as well 🫠
It will also give us radical abundancy so who gives a shit? It's like everyone read chapter one and put the book down.
I'm personally heavily invested in that robotic future. Yes it'll disrupt employment as we understand it today... But I prefer cars to beating some horse with a whip when visiting family. Not all loss of employment is bad... Even if a lot of those horses ended up as dog food during the transition.
This disruption is the whole point, and it's one of the reasons that despite the red herring that "AI is coming to destroy", the progress on it continues. Besides, the structure of all our modern paradigms are all bunk, and what you don't realize that one bunk system is just getting replaced with another.
Thanks to the ignorance of the millions upon millions of layman out here, our institution are only so brittle that they will crumble by such rapid change in advancement because people of ignorance allow such corruption to perpetuate.
So despite any perceived activism towards the abolishment of AI, people should remember that their inaction against inequity is also part of the problem.
13:58 "the number of amputees is expected to double by 2050", what are you not telling us Dagogo? What is going to happen in 2050?!
more obesity, more diabetics getting limb amputations
More population
@@tuvaaq…..Must be in USA….
sweatshop and warehouse accidents
same. wild ai foresight i guess.
550 million amputees ???????????!!!!!!!!!!!!!???????????? that can NOT be right !!!! : 8 billion people, 550 million amputees: 1 person on 16 ???? ABSOLUTELY NOT !!!!!
Thing is, in these 8 billion there a lot of people from underdeveloped countries with a lot of manual labour with close to no automation, so a lot of amputees due to sickness or accidents. Still not sure about the number, might be an overstatement
This is a mistake, I believe he meant 55 million- I’m an amputee and the sources I trust are all in the 50- 60 million range
55 million maybe. no way 550 million.
Yeah more like 55 million worldwide, though the rest of that statement rings true about the projected numbers in the US. All the obese diabetics that end up with limb amputations from shovelling high carb high fat 💩into their bodies everyday whilst watching Fox News...
you try operating a sudanese steel mill with your hands
Thanks!
Pharmaceitical companies cutting down on drug development time is only going to benefit the shareholders. It will not make drugs cheaper or actually help people.
As someone who has studied biotech I can assure you it will benefit people cause drugs can be sent out faster.
Like imagine a bacterial disease as infective as COVID-19, and instead of waiting 5 years to develop drugs, you can do it in a matter of months or less
Also only a problem in America with it's f**cked uo insurance
You're completely wrong drugs will be a lot cheaper because it takes less money to build it using AI which means they can sell it for cheaper cuz they have less overhead so they make more profit. And if one company doesn't another company will that's how the free market works I build it for cheaper other companies make their product cheaper
@@EspeonMistress00 I just had to pay $4,000 for one bottle of medicine because there was no generic alternative and my insurance screwed up. How is this even remotely justifiable? Drug companies will simply hide the vast savings that "AI" brings and pocket the profits. Nothing will change. They are evil incarnate. Ask all the victims of the Sackler Family.
@@EspeonMistress00the only way this is a benefit is if drugs are created and sent out faster while all drugs also come down in price significantly.
Man, I remember back around 2018 I bought a 1080ti off of a doctor who was experimenting with deep learning. He used the GPU he sold me to scan XRays and MRIs. He explained the whole process to me. It was mind blowing, look how far we've come in just a few years :)
Is there a reason why Coldfusion is basically the only TH-cam video series that's promoted by Bing news?
Cold Fusion is a complete joke now and for the past 7+ years. Highly compromised.
@@fluxonite Can you tell me what you’d like to see improved?
Cause he's amazing that's why ❤
@@fluxonite oi why not reply, do you not have anything actually useful to say
Your introduction of "Hi welcome to another video of cold fusion "is soo underrated, It gives me goosebumps
As a researcher myself, I strongly support the use of AI to tackle hard problems that can benefit humanity. Rehabilitation robots, robots used in extreme conditions and others could benefit heavily with AIs. I do not support the use of AI in creative jobs. AI should be used as a way to help humanity in the long run.
As a musician and painter, the term "luddite" comes to mind. AI is far and away THE most creative tool ever devised and those who try to hide behind some mythological concept of the human "more", call it a soul, call it a spirit, call it whatever you please as special pleading is special pleading no matter what the term you use to obfuscate it, those people? They'll be left behind in history's waste bin.
Good thing this kind of AI doesn't function without a consistent feed of immense cooling and electricity. Oh but don't be luddite when you be come to expensive to hang around in the lab@matthewlawton9241.
@@matthewlawton9241 I don't think you need to qualify. People? They'll be left in history's waste bin.
Stringent copyright laws around AI is exactly what will bring forth the corporate dominated AI, as only the megacorps will have the money to buy the million licenses for every piece used in the training data. Taxing the AI companies and using the money to enact UBI and then embracing open source is the better approach imo, than throwing sticks in the wheels of AI, creating massive barriers for entry in hopes that artists can hobble along a few more years. It's not like the royalties off of that 0.00001% training data contribution is going to net you a living anyway. The commission will be a few dollars per piece if that.
@@MidWitPride Let's be real though, AI will become corporate dominated no matter what. It's like setting up an ISP-business, the cost of entry is far too high for any small actor to compete. It costs hundreds of millions to do a training run on a frontier model. It's not like software, where a small team with a good idea can revolutionize the industry(like whatsapp), small players just don't have the compute to compete.
Open source doesn't matter all that much either, as it is far from enough for people to duplicate the models(GPT2 is open source, it's not extremely complex, honestly. but the training needed is just prohibitively expensive). Ironically open source might benefit the big guys, as if for some reason a small lab got blessed by god with a fantastic idea, and trained a model that actually made novel progress, the big players could just instantly copy the architecture, and iterate on it with their far greater access to compute. The barrier of entry that is compute is so much higher than anything copyright could ever pose, that it isn't even funny.
The only way AI doesn't become corporate is if the big labs are nationalized. AI under a capitalist system is guaranteed to become corporatised.
The current rush to A.I by every business imaginable reminds me of the early days of the Internet when every company had to suddenly have a website. They are all so vague about what their A.I. is that i have no idea what they mean by it.
I am no fan of AI. I am retired and have no use for it unless for better or at least quicker video and photo editing. My son is a graduate degreed computer engineer and code writer and he has no fear for his job or future. I have no fear at the moment, but I fear for my grandkids, especially with our government already out of control. The next war may not include weapons other than supercomputing AI. I can see computers/ robots taking over most everything at some point, with or without human control.
> I can see computers/ robots taking over most everything at some point, with or without human control.
My best guess, all the humans die shortly after the AI takes over. (Ie without humans in control)
von wrote, _"My son is a graduate degreed computer engineer and code writer and he has no fear for his job or future."_
Then he, or you, don't understand what's unfolding in real time. I don't mean that as a disparagement to either of you, but the rapid pace at which AI continues to progress is likely incomprehensible even among the smartest segment of the population. Part of the problem here is simple hubris: humans tend to think of themselves as the pinnacle of intelligence but this simply isn't the case. We're a lot stupider than we care to admit, and I say that as someone who from my teen years to my now retirement years, has tested quite highly on IQ tests. That's not a humble-brag, it's just offered for context that I'm not exactly some ego-driven, dopamine-seeking yahoo with a baseless opinion.
In my view, the biggest challenges associated with AI have less to do with the technology itself, and more to do with the people who control its implementations. 'They' will continue to do as they've been doing for millennia, and those of us who have less agency in the matter will become unwitting victims of poor *HUMAN* choices, not directly of poor AI choices. We have met the enemy, and they are us -- by which I mean: they are human.
Thousands of years of written philosophy instructing us in how to delve deeply into important matters, and ultimately it's those with the greatest deficit in empathy who rise to the top. This is an indictment not only of such individuals, but also of the collective that fails to recognize this dynamic; or, if they do recognize it, who then fails to prohibit it; because they have their own narrow-minded self-interests and cannot be bothered to understand the truth of reality -- as long as they get what they're after, Rome can burn and they'll be focused on manufacturing violins for the masses.
@@RichardHarlos Few people want to destroy the world, so if the world is destroyed by AI, it will probably be by accident.
Our understanding of what code leads to what AI behaviors is rather rudimentary.
@@donaldhobson8873 wrote, _"Few people want to destroy the world"_
Ignorance is no excuse. People don't care whether what they do has long-term consequences. So, when you say that few want to destroy it, that's not nuanced enough to portray the truth of the matter.
Which is why _"it will probably be by accident"_ is both true, and yet incomplete. The precautionary principle states that in the face of an unknown, one ought to err on the side of caution. AI is a vast unknown, so to race down that path as if we know where it leads when we don't, and in fact we can't... is folly... hubris... ambition... none of which give one whit about the future.
So, while it may be true that they don't consciously, deliberately want to destroy the world... it's also true that they don't care whether they do. And that's a problem.
@@RichardHarlos Agreed.
luv the new intro!
The problem with AI is the huge amount of data required for its development is stolen. It's not just image generation or LLMs. All image and language recognition AI is made using stolen date. Tesle probably gets driving data same way. Google gets all their data by stealing it, there is music too, I'm sure I'm missing other fields. The companies are getting rich with this stolen data. They say it is opensource but use it in their own tech and charge people for it. Then there are cases like adobe that make people sign an agreement to use all the data made using their software which the users had to pay for in the first place. When the industrial revolution was starting people wondered what they would do with all the free time and abundance of resource with the use of machines. None of that happened, the rich just got richer. If any part of the AI required stealing data directly or indirectly then the people should be compensated fairly, or it should be FOSS and cheap, companies shouldn't be allowed to get rich off of it. People are already losing jobs. If it goes on like this then more people will be poor and they won't even be able to afford all these new AI innovations.
I find this argument lackluster, because same if it is not stolen data, people are still against it.
Why is stealing data bad?? If we can steal data and cure cancer in one day versus not steal data and take 100 years to solve, what is more ethical??
You stole all the data used to train your neural network too. We all did
So refreshing having genuine science and tech channels like this, glad to be a long time viewer. 😊
Years ago AI was called 'Expert System'. CPU given a database of a 'subject'. You would submit a query, it would search that database & kick out responses.
Then as now it really matters who designs & maintains that AI with the data it uses.
It has access to the full internet, OK how many times have you done a search and found conflicting if not out right contradictory information?
OK how will that AI deal with that? Oh trusted sites, OK is that 'trusted' site bias in some way???
The algorithm for fact-checking is always the same how many "top" sources are saying something? So unless there's a lot of "top" sources saying the same misinformation, it will usually be able to figure out what's true. But even if it doesn't you can't expect anything to 100% perfect.
Datasets that current frontier models are trained on is highly refined for quality, as the quality matters A LOT in intitial training. The data is not the same as the garbled mess a web search would get you
I always like your take on these new technologies. The kelp forest thing was my favorite part, although prosthetics and new medicines are right up there too. As always, I love your presentation style. Thanks for the show! PS: Watched this on Nebula (as always) but came back here to comment.
"Al isn't coming for human jobs" Personally, I think people should stop saying this, even as a joke or with the "yet" on it. There's no value in that narrative. It WILL take a majority of jobs, possibly by the end of this decade, and we need to start preparing for that eventuality, right NOW! I lost my career path to Al last winter, and I know half a dozen other people who've also lost jobs to Al since last year! This IS coming, and the longer we delay adapting our society and economy, the more unfair it will get. but for my robust investment portfolio that I'd built over the years with my financial advisor, Abby Joseph Cohen, I'd be hungry and homeless
Abby Joseph Cohen Services has really set the standard for others to follow, we love her here in Canada 🇨🇦 as she has been really helpful and changed lots of life's
The very first time we tried, we invested $7000 and after a week, we received $9500. That really helped us a lot to pay up our bills.
I know this FA, Abby Joseph Cohen Services but only by her reputation at Goldman Sachs; even though she's now involved in managing portfolios and providing investmnt guidance to clients. I have been trying to get in contact since I watched her interview on WSJ last month
@@VonNothiasWell her name is 'ABBY JOSEPH COHEN SERVICES'. Just research the name. You'd find necessary details to work with a correspondence to set up an appointment.
Great share! Out of curiosity, I searched Abby Joseph Cohen Services online, found her consulting page ranked at the top, and scheduled a call session. I've read many reviews about advisors, but none seem as impressive as hers.
Nice to see you “in the field” doing investigative reporting! ❤
Anastasiaintech reference, nice
while replacing a graphic designer with AI probably is quicker and probably does save money. this approach leads to a lowering in quality that’s hard to measure and hard to address from the top down. New ideas have a hard stop after the person has left, all you get after that is variation. which can work for a while and in a limited way but fails everywhere else.
The negative perception of AI exists because every company is slapping "Powered by AI" on literally everything. I mean, we’re one step away from finding a light bulb with AI that can "predict" when you're about to flick the switch! Next thing you know, your toaster will be "Powered by AI" and trying to give you life advice along with your toast.
They do this with everything. After Apple popularized smartphones everything became smart. But is basically meant the given equipment, tool, device can connect to the internet. Heck, we even have smart washing powder.
Nowadays I see a big rise in protein. Everything has protein in it, and they act like it's the holy grail of a fit body. "Do you want to eat a chocolate bar, this one has protein in it, so yoi don't even have to workout!".
Also the electrification of every vehicle. I saw electric skateboards, rollerskates, beside the more common cars, bycicles, and scooters.
Everything is smart, has protein, can transport you with using electricity as fuel and of course powered by AI.
@@nupagagyi You my man is absolutely on point.
Very interesting Video...
Good job Mr Dagogo..👍🏽✍️ Thanks a lot..🤝🏽
Man that kelp restoration project sounds like it really, really doesn't need ai.
Of course it doesn't. It's like when crypto bros feel your business "needs a blockchain Bro'." 🤣 Ask yourself what you actually took away from this video besides misrepresented information and empty platitudes. 🤣
It does need an automated way to detect patches of kelp on thousands of square kilometers of water
Thanks for the good news for a change. The kelp story was awesome.
0:14 trashy? Woah woah woah!! Shots fired 😂
Because it's overhyped
@@dariuspalmer2829thing is, artists are ok with most other uses of AI. Just not for art purposes.
5:41 the lithography optimisation isn’t really AI though I suspect. I’d would expect it’s using mathematical optimisation algorithms, rather than neural nets or similar AI concepts, where the individual calculations in the optimisation algorithms are parallelised on the many GPU cores.
if corporate greed stays out of it.... thats not a dream. it will take a miracle.
Or force.
3:50 Anastasiiii!!👏🏾 She's pretty great. Happy to see the reference 😌❤️
Graph showing performance jump by nvidia generation at around 5:47 is due to support for smaller data types, eg fp8, and fp4. The reason it gets so much faster is largely memory access, there is less data that needs to be accessed/moved around, so it’s fundamentally a bit of a misleading graph.
Thanks to the ColdFusion team. We all needed a more diverse view other than focusing on creators.
The problem has never been AI, it has always been capitalism and "corporate greed". Capitalism creates monopolies, it is in its DNA, and AI is an oportunity to create a monopoly to rule them all.
Not a perfect system but the best one we have
The issue is human pathology.
@@TechnicalsMatt- It's the only system we're "allowed" to have.
@@fluxonite Allowed by who? Because there's definately countries that have tried other systems. Many of them are either in the process of collapsing, or have already done so. Even China is becoming increasingly capitalistic in the shadows, even as stoke nationalism by promoting their fake communism.
@@Jo1162 We know that's not true because marketing is a thing, and has been provably done to the point of deception. Not to mention buying up of any would-be competition.
Thank you so so much for making this kind of video covering the parts that get glossed over all the time. It's exhausting seeing constant bad faith and exploitation and just a complete lack of acknowledgement for people working hard to try to _actually_ make life better for people with the new technology that previously we simply didn't have a means to do.
i had forgotten about nvidia's culitho
spanish speakers find this name amusing
As a Hispanic latino, It is very funny indeed
wat does it mean?
@@stereo-soulsoundsystem5070 Culito in Spanish means little ass
@@stereo-soulsoundsystem5070 just search the word "culito" and go to images
humanity is about to hit another peak with technology through A.I. but we haven’t yet seen how this is going to work out in the long run inside our current system and society
Agreed however its catastrophic for the internet
If we can turn this planet around and overcome our dark side . . . . . . there is an amazing future available.
hi Dogogo, i am a follower of your channel already for a while. Thank you again for making sucvh videos. I miss one essential discussion here: All technological improvements are bound by the ethical political decisions that are in the hands of those that can control the output of such technologies. It is not the technology per se, but for what purpose it is used. Eeven making a Paracelsus analogy, the diference of cure or poison is not the substance but the dosage. On this terms, i believe that most of us are skeptical with the bigger decisions regardings those technologies, as we have seen in human history, a.k.a atomic energy.
"New medicines with less side-effects for a cheaper price" - These companies have patents and they are not beholden to the needs of the customer, only the shareholder. As someone that works in software (which is being BOMBED by AI), it is clear that most people will not profit from AI. There will be more software and it will be worse and it will not be written by humans. It will cost the same and the only people profiting are shareholders. No technology ever has actually decreased the amount of work we have to do and made life less stressful it will only squeeze more out of the workers for the profit of the capital.
Hey man I know you guys are excited about this and all, but I just want to put down a wet towel over your heads, LLMs and classical neural networks are not the same, a bit similar in how they run but NOT the same, LLMs are only useful for *certain* tasks, one of them certainly is NOT prosthetics, so mixing them up in a video is misleading at best.
Why do I say that? because if you take out LLMs and what they are useful for, which is basically prediction engines that are fairly slow in execution (ideally for text where the amount of possibilities is less), prosthetics are not fit for that, you require in real time movement, for that you have to use either direct nerve connections or a sort of real time AI, ideally, a very very good near AGI AI because there are inherent risks to it.
If we assume that, then, there has been no actual "lightning speed" porgress to it, it's been the same "slow and steady" AI progress, which while it is somewhat happening, sure, it's definitely not gonna come out tomorrow.
LLM investing companies outside of Google have near to nothing going on regarding non LLM research and as such as actually doing 0 progress towards artificial general intelligence, which will NOT be an LLM, LLMs will NOT be AGI do not let scammers trick you, read up how they work, read the papers, use them, the most up to date models in real world scenarios, listen to the head of meta AI talk to Lex Friedman about it as well, you will quickly form a conclussion that the space is selling a bunch of vaporware that doesn't exist.
When it comes to LLMs, AGI doesn't have to be sentient, it has to be useful. And it turns out, language is a way of storing ideas, that lets them be manipulated, and text works fine as a method of output.
Honestly, we are closest to reaching AGI with LLM's, even if it doesn't look like you imagine, and even if it's slow, that doesn't mean it isn't useful. For example, where it can write an entire programming project based on specs, which we are pretty close to. It's flawed at the moment, but with better models and training, there is nothing in that which couldn't be done by an LLM.
Quite frankly, it doesn't matter if it actually "Understands" or if it's just a Chinese room. Who cares? If it is able to produce quality results, that's all that should matter. LLM's are able to act like they understand, and that's quite good enough.
And I'm aware of how they work, it's my professional field. I have probably read more papers than you have on this. Just because someone arrives at a different conclusion doesn't mean they know less than you.
(Also, he never said LLMs powered prosthetics, so I don't know where you are getting that.)
You don’t need agi to create a disruptive technology utilizing ai. Agi is a world changer, absolutely, but the world could change with a highly useful ai system regardless of its agi or sgi.
Someone only reacted to the title and didn't watch the video....none of these shown are language models. He mentioned at the start that he considered "generative A.I" to be a different topic from what this video was covering. "Google have near to nothing going on regarding non-LLM research..." Did you forget about Google Deepmind and their AlphaGo model that helped kick off the A.I hype before LLMs showed up?
Amazing video, thank you for making this and pooling the sources. Wish there were more videos on this topic. "Other side of AI" type stuff.
It's annoying when I bring up AI and ppl instantly have an opinion on ChatGPT and etc... I'd say that these conversational LLMs are still pretty stupid. Literally these current LLMs don't have "logic", they base answers on weights. Literally guessing what you want hear based on the weights (yes and other parameters). I wish there were more videos that discuss AI being applied in science and tech fields and not chat bots lol.
I have Chat GPT 4o and right before watching this video, i had 2 hours of conversation with it. Compared to talking with friends, it was way more most insightful and productive. At this point, it is not like talking to a human, it is better.
Productivity and Insightfulness is usually not the objective in a conversation with friends.
@@jean-claudelemieux8877 I couldn't agree with you more, Jean Claude. Nevertheless, may I point out that you are rightly considering the word "friend" in the deepest human sense of the term? (I think that IS what sushantkarki seems to have missed with his apparently reckless, yet otherwise curious comment, wouldn't you agree?). Goes without saying: I really like your vision of true human friendship (Cheers!! :)
@@edwardenglishonline
I do not believe there is something to say about true human friendship here. I was simply using the, in my opinion, common sense of what is a friend.
Talk about the word “delve” all of sudden being a top 3 word used since chatgpt came out
The "Hype" also leads to significant investments from the biggest companies in the world. The AI haters are the equivalent of the people back in the day who said computers are all a waste of money or cars will never be a viable form of transportation. I don't get it, the potential of this technology is limitless.
no, iron fist era is coming - AI main task is surveillance, censorship, compliance monitoring and enforcement. have a look at what Oracle is doing and what all agencies like WEF and govts are legislating
You also need to consider how stupid silicon valley appeared after the Sam Bankman-fried fiasco
I remember people back in the NFT peak hype cycle had similar viewpoint and really any Elon musk fan after he announces one of his fantasies
What a completely saccharine naive perspective. 🤣 Yes I'm sure "the best companies in the world" plan to usher in a utopia for all of humanity! "Think of all the possibilities!" 😉
What a great start in the morning seeing an 80k coldfusion sponsorship! 💯
Bro never miss
Dr Suneel Dand is great and i am happy you featured him!
I'm a Gen-X'r and 59. My background is in programming. Since the early 80's I've known that technology would overtake many jobs done by humans. Civilization is going to advance. Unfortunately there are those that refuse to live in the future.
The problem is capitalism.
@@Julez60 Stop spamming the comments section, you're not going to convince anyone by nagging them.
@@kiefermattern917 so butthurt lmao cry about it
@@Julez60 Ok bro.
@@Julez60We're going to have to live with it until a decent alternative comes around.
Dagogo, your presentation is highly relatable. Thanks💙
It's sad that a primarily technical TH-cam channel has to make a video defending the premise that AI can be a good thing.
So many people piled on the hate train for something that's going to hit every aspect of our lives, hard, and somehow are blind to what's just on the horizon.
Name a single positive aspect for those not in the 1%.
You will not be seeing "trickle down AI."
yes people are so emotional and in denial. I have mostly hidden comments on youtube and even tech websites because people seem to lack any sense of how big the shift will be eventually and are just clinging to any examples they can of LLMs failing
Another excellent video essay. Thank you for sharing!
It surprises me that no one has mentioned "Taguchi" in the context of AI.
Top-notch coverage about the bright side of AI.
Very well done, Dagogo!