It can automate the shit outta your job and pass it other sub processes to run the random shite. It sees your hate as the problem to solve. P.s: thats the ideal AI anyway, probably take us 5 years or more
My company is producing "AI software" for the NHS. Its not AI. Its an automation algorithm using SQL that could have been done a decade ago if people coulr have been bothered. But hey, its AI now!!! Buy our AI!!!
@@TheRealHassan789 You cannot make detail out of nothing. AI will do it, but it will just be a hallucination, which is absolutely pointless. Actually, it's a great metaphor for how useless AI is.
@@d0m96 check this paper out: “Personalized Restoration via Dual-Pivot Tuning” … it’s pretty cool. Because it lets you bias the “noise” with known facts. Hence AI can actually do a pretty good job.
So when they need the content the social contract is that it's free use, but when it comes to torrenting tv shows and movies it's like stealing a car? 😂
Also ignoring that free use implies that the content is being used in either a transformative manner or for education purposes. Using it as training data for a commercial product should not count, as the content is not transformed in any way by the actual people handling it, or selling the product it was used to train. In fact, only the end user should be protected, as they are then using the data transformatively to generate things.
I work in a corporate office for an international company. As an IT support manager, all my leadership keeps asking "cant we automate this with AI" ... they say this about everything. My plan is to stay human, analog and manual. In the future, that will be the specialty..
Those same people two years ago, when asked if there was a budget to create traditional automation tools (which most of these "AI" products are anyway), would claim it wasn't worth it.
The amazing thing is that all the tech bros and young keyboard warriors support this sort of statement because they completely do not understand the implications.
Everything is free for the taking regardless of laws, ethics and common sense but don't you DARE take away from them or they will run after you. That's the average way of thinking of these people.
Too many companies are claiming to be AI when they are not . Its become a buzzword to be accepted by places with nothing more than a normal software algorithm.
A friend wants to remove 90% of the devs from his office because AI can now generate code. He wants to keep a few to manage the code. What he doesn't realise yet, is that although useful for some code ideas and snippets, a lot of what it churns out is bollocks and you have to correct the AI and make it do it again. Then you have to figure out how to add that code into your own. We are far away from replacing humans in some roles, but people really think that AI is better than it is right now.
I find it useful as a better version of smart complete; but you still need to know what you want it to do, how to architect a larger system and proof read what ever it spits out.
I used to work for a company making packaging machinery, our customers would always insist that we should have the latest controllers, use the latest motor technology and have lovely HMI's everywhere, all of our competitors did.... this was all nice and good, we could do all of these things but at an upcost to the customer. I explained to many customers that our standard machine offerings with their basic control systems were proven, worked well, were easy to operate and maintain and cheaper than adding the technology they wanted, our standard machines also had a much quicker turn round and we made more money on them. The question I ended up asking our management was do, we really want all this technology on our machines if it's not really that beneficial, is more complex and is costing us more? The answer was that we need to use this technology because all of our competitors do, we can't be left behind... perhaps that's the issue with AI we have it because we have to.
Calling these LLMs "AI" is a stretch imo. I am surprised that this hype even caught on since these were a thing for quite some time, and were never ever envisioned to be extremely groundbreaking and paradigm shifting ever. At least nowhere near the potential the advent of widespread internet coverage had.
It is ridiculous how many promises are made regarding AI generating text, music, code, pics videos and ultimately be smarter than humans. And all it is is a glorified optimization algorithm that finds a most likely word, note, token, picture or video fragment that would fit the best given the input requirements and limitations. Sure it will be able to drive cars anywhere in the world _someday_. Or deliver pizza. Or perform surgery. But the current AI will never be able to invent.
i recently did an interview with a startup company in germany that is using ai to attempt to automate trucking. the job is working at a desk with a gaming wheel and watching 4 trucks do their work at a port, when the truck gets stuck you take over and help them out. meaning a single driver is technically driving 4 trucks at once. i think most ai will end up in a similar capacity, reducing the need for labour. otherwise the entire planet will grind to a halt because a bird flew in front of a camera.
So, you could rob a truck without the need for hurting anyone? And you could strip them for parts too? Cyberpunk esque call the police when danger is detected sensors, or other defences could be easily circumvented with the right knowledge. So robbing trucks
What happens if TWO trucks get stuck at the same time? What happens if three or 4 get stuck? Is the port going to allow the company time to play with their legos, and stall the whole operation? How many truckers salaries do they think a full hour of a whole port not operating is woth?
I agree with the points raised here. If this technology could be used to complement truck drivers, it would be fantastic. Imagine a situation where a driver falls asleep at the wheel, and the system triggers an alert, allowing a virtual assistant-enhanced by advanced tech but still human-led-to intervene and prevent a potential tragedy. However, the real issue is that some companies are more focused on cutting labor costs and maximizing profits rather than enhancing safety and support.
@@TheManinBlack9054You’re forgetting about the consumer. Christ, who’s buying the goods if AI has replaced said people in those jobs. Ten points for you being an absolute genius
No, that's actually the most terrifying aspect about it. If nobody has any money and become dependent on "UBI" to live, we will devolve into a dictatorship rather quickly, as the goverment (and their friends) will have 100% control over your life. I think that is the whole point actually, but that is another conversation.
It's clear the Gareth Southgate AI videos are of the highest quality order and have reached General AI. They could probably come up with a better strategy than the real man.
The problem with a lot of A.I. as I see it now is the amount of time, money and effort that is being put into replacing people and doing general tasks like video creation, chat bots and investment tools. The system is logic-based and should be geared towards managing and improving things like urban planning, resource management and advanced robotics.
It's easy to excuse cases where AI messes up generating things like images and videos but giving AI complex tasks could cause catastrophic results if it messes up.
@@rayauxey I get where you are coming from but that just proves further that it isn't yet fit for purpose if people are afraid to, or unable to use it where it would be of most benefit !!
The tasks that "AI" would be good at is managing efficient workflows. But wait...thats Business Process Automation (BPA) and that's been around for many years.
Conventional ML techniques work better for those kinds of things as they're orders of magnitude more efficient and accurate. LLMs are mostly a very compute heavy way to generate text (or images).
I agree generally but S&P 500 is very top heavy with these handful of AI hype companies like MS, Meta, etc. diluting the diversity appeal of a S&P 500 fund. They do have solid underlying business, but they will take a hit along with S&P 500 funds when AI bubble pops.
Look at the top 7 holdings of SPY and you'll see that they're all AI related plays. (i.e. MSFT, GOOGL, GOOG, META, AAPL, NVDA, AMZN). Collectively they make up 31+%. And that doesn't include other names that may have been pumped up by the AI hype also in the index (i.e. smaller semi companies, data center reits, etc.). In other words, thinking you're diversified holding a S&P 500 index fund and won't suffer from the fallout of an AI crash is fantasy. S&P500 has basically become somewhat of a AI theme momentum fund.
Google AI told me that a conversion from 107kg to Stones and Pounds was over 31stone (It's actually 16st 12lbs). It even laid out the steps as to how it got to the answer. Problem there was in the explanation it arrived at a figure of 8st 12lbs! But still told me it was over 31st. I've also had to get rid of Google Gemini and revert to Google assistant as it just kept getting answers wrong.
@@carultch Yes, but they put their AI (Gemini) on Android phones as default, to replace Google Assistant. After this incident however, I have reverted to Assistant.
*IT APPEARS WE HAVE REACHED PEAK TECH* as someone said "“we perfected the internal combustion engine in about 1940 - since then it has just been minor incremental improvements - 80 years of tiny advancements - maybe we have reached peak tech"
@@adamaglionby7438 Actually James Watt's engine was an atmospheric engine but with incremental changes to Newcomen's invention, yielding impressive gains in efficiency. The first true steam invention was I believe down to Trevithick, in about 1801. I have a soft spot for Trevithick: he won a bet on his steam-driven "road" engine ("Puffing Devil") being able to move coal near Merthyr Tydfil better than pit ponies. But he was so happy to have won that he retired to the pub with his mates, leaving the engine puffing away, as a result of which it then blew up.
This is why Microsoft now want access to everything you do on your computer. The low hanging fruit on the internet is gone so now they will scrape your every minute of life.
I hope your correct and there are no emergent properties from this investment bubble, because with bugs it would be dangerous if it did continue to be more advanced while profit driven, for something all controlling and powerful there can be no bug ever or its all over. Luckily there is SSI with Ilya Sutskever, machines creating better machines in a closed loop are a powerful thing. But with no product ever, only one final release, it's investors are investing in 100% hype.
The AI in Google search isn't useful at the moment, but it is learning from everyone that clicks on an AI result. We are teaching it, correcting it, improving it. At the moment it says 'oh, you are quite correct to point out my error. I am still new and may not get everything right just yet'. In a year it will say 'Sorry, but my answer is correct - you are in error. Please wait while I change all references on the internet and erase all your personal information according to the Computer Heretic act of 2026.
This is another AI grift. Companies do use user interaction and feedback as signals to train and optimize their models, but it's happening with A LOT of engineering effort, and it's just a small part of the overall work. It's not even close to "the AI is learning from you", nowhere close.
Google doesn't give a shit if the answers it gives are incorrect. More inquiries = good for business, even if it means their service is poorer in quality than it was before. All they care about is how much time you stick around taking in ads.
AI chatbots are where a lazy human asks a lazy robot for answers. The answer is usually wrong or incomplete, so the human spends an hour chatting to it, trying to make it understand where it went wrong and apologise.
Sadly, typing "understand where you went wrong and apologise" does the same thing as that hour of chat It still won't understand, and it isn't really sorry - but it will give you another dubious answer.
You make a good point about the sources required for AI to continue to make gains. This is probably why Adobe tried to muk with their terms of service to allow them to utilize their customers photos to train their model(s).
Adobe’s AI generation is pretty horrible too. Probably the only useful improvement is generative fill for removing elements in an image or extending edges.
Even if the scaling laws are correct, nobody is going to be convinced by tech bros saying "please gimme 1 trillion to build 3 nuclear power plants and a teradata center, the result is guaranteed AGI!!!" The problem is, we need those trillions to do other things now, like heating, air conditioning, building houses, all that boring stuff.
Ah the smartphone analogy - bought one in 2020 for 200, now a new one of the same type is selling for 250 and is still competitive, the only upgrade I consider is a new memory card that doubled or quadrupled in the meantime
Wait, all content on the internet is free use and that has been generally agreed upon since the 90s? And here I have been diligently paying for my movies, games and software for years...like a fool. That CEO was literally saying that piracy is ok.
Reasonable point of view. I think the problem with AI right now is that the whole market is inflated by companies that don't generate actual revenue with the tech (at the moment). Everyone is investing billions but these investments don't translate into actual useful products that can be sold. But I'm not sure it's as obvious as you said because I have no idea how long it will take to create functioning products. Do you think that things like tesla robo taxi aren't realistic in the near term?
Correction: GENERATIVE AI has become bubble. Regular AI is doing fine - finding cancers, diagnosig patients, doing optimisations and defectoscopy....3:51 - same about Ukrainian War or any other subject realy....
I work in IT leadership role and I have an innovation team looking at AI more specifically LLM and I can’t find a single commercially viable use case other than chat bot even then it’s not that commercial. I also recently completed an AI course at Stanford and looking at the latest in AI it’s actually very far away for anything to be fully automated. We’re talking another 10 years at minimum.
Yea advertising, entertainment, automation, heatlcare, Agriculture, weather forecast and so on but you only think about chat bots. You're the average IT boss, useless and years behind.
@@drtomgao "We estimate that generative AI could increase labor productivity by 0.1% to 0.6% annually through 2040. Combined with other technologies, work automation could add an extra 0.2 to 3.3 percentage points to productivity growth." says McKinsey. I personally think ai assistance in costumer support will absolutely take over, it's a way better user experience than having to wait 10 minute just to talk to some Indian guy. Entertainment is gonna be crazy, there's already weirdos who date some ai now imagine if there was a decent actual ai attached to a robot, millions of people would buy that.
@@Torbintime ok so based on your comment you're just making assumption based other people's view and comments. You haven't received direct education and looked at the maths behind LLM nor understand cost of computation of inference/prompts nor understand costs of IT/AI projects and governance of AI hallucination risks. Lots of unfounded assumptions not based on a single tangible example of a commercially viable usecase in 2024. I can also quote lots of benefit for AI in 2070 but we're talking about right now and the next 5 years.
@@Torbintime those are very modest expectations for a bubble valuated in the order of HUNDREDS of BILLIONS of dollars. AI being a bubble does not mean that AI itself is useless, just that the expectation is vastly out of all proportion. with that kind of money, if AI fails to replace literally all white collar jobs by, uh, next year, it's going to fail catastrophically. literally OpenAI and Anthropic are running out of outlets and are starting to scrape the bottom of the barrel, starting with the saudis. When you see sam altman begging softbank for money, you better cash out.
@@daveabc12 Scottish Widows Active Money Market fund. It's the only one available in my company pension. This is where new contributions are going. Previously invested money is staying in passive global equity index tracker and over five year inflation linked Gilt tracker funds.
Often a good idea isn't self sufficient, but dependent on other parameters. Internet, computing power and the cost of computing power: when that matured, LLM's became opportunistic, and it's a magic wand. We shouldn't be angry at a dog who doesn't pronounce all English words correctly.
i have no idea why this bubble wasn't even created when the tech was already being developed now they just rebranded the same old machine learning and neural networks as AI and now everyone is investing
It's LLMs and text2image and recently you see it more accessible easy to use and in the hands of the general public mainly but the typical consumer/investor doesn't know that, they only know the general "AI" buzzword which is all that really matters from a marketing standpoint
Because major part of stock ownership are USA citizens, and in USA every second MacDonald's employee buys stocks. An investor mob like this is a blind trend follower. USA top money people did a good job of getting this crowd onboard, much easier to fleece compared to specialized investor.
5 - 10 years from now whenever you call or send a chat/text message to your bank, phone company, utility company etc. you will talk to AI instead of a person. If you complain enough and ask enough times it will transfer you to a real human, but most likely it will answer your questions better and faster than a massive call center of people who don’t really want to be there can or ever will. This transition is already in progress. Advances in robotics opens up manufacturing, construction, farming, food prep, transportation, repair and maintenance work to AI as well. Basically every blue collar job will eventually be done or assisted by robots running on AI. AI and robots will help us get through the population collapse which is currently in progress all across the developed world. It will allow us to do more with fewer people. It won’t result in no one having a job. People will still have jobs, but they’ll be different than they are now.
@@EamonnMurphy-cc1dt autonomous cars are not AI same as all other robots just preprogrammed following specific instruction sets based on "IF this happens DO that"
Often, we have the truth right in front of us and either fail to see it or choose to ignore it. Thank you for your efforts in spreading this important truth about the AI craze.
Sasha is absolutely correct. This is a massive bubble. Curent AI is just a fancy data fitting method. What will be truly revolutionary is AGI. And we are as far from AGI as mammoths from social networks. I am not even sure if AGI is possible. It would take a completely new approach and understanding of consciousness and stuff. AI has peaked. AI will not get significantly better than it is now for at least hundreds of years.
This 100%! One thing that makes humans human, is this collective socially constructed lie that symbols (words, pictures, sounds, etc.) mean something and aren’t just arbitrary connections. AI simply can’t lie. It can tell falsehoods (purposely programmed or by accident), but it can’t inherently live in our social lies.
AGI is also not a magic bullet. Creating AGI simply means creating another sentient being. What's the guarantee that AGI will even want to do chores for humans
Ok I laughed so freakin hard at that AI response to the question "how many rocks should i eat per say". Thank you for that Sasha, after watching the debates I seriously needed a good laugh.
Nonsense can big data, Draw pictures, write poems, translate between languages, speak naturally, understand humans, code, make music, operate robots...
What a great video. I find it funny that all of a sudden, all these companies have AI. How did they get to it? It is amazing. However, if you use Google Maps and use to get directions, the thing is so bad, that at this second, I am on I4 and if I am in traffic and 508 passes over me, I am magically on 508 and get a new route on how to get to my destination. Unbelievable. Or, that Google shows me a traffic jam ahead because there are a bunch of cyclists standing on the side, or that it gives me a route through a private property, or a route through a dirt road if there clearly is a better road, or 2 other suggested routes that make no sense at all. I always wonder how these valuations come to live if companies are clearly so bad in day to day things and with this AI stuff, it gets even worse.
Less than More 10 trillion dollar valuation. Without making single dollar in net Income. I guess we don't know better. Certainly smarter people than us are r this show.
I would love to see AI working at traffic lights so when I’m sitting there at a red light watching a green light with absolutely no bloody cars wouldn’t happen.
We've had camera pickups in traffic lights for over two decades. Whether they're widely deployed anymore, I don't know. My father used to install them for local government.
Why do you need AI for that? AI implies ability to learn from new data and change its behaviour. All you need is a normal script with presence sensors to detect vehicles waiting in a lane but not in the other. Such system is already implemented in some places since years, although rarely. Usually it's crosses between very tiny roads with negligible traffic crossing important veins that do not need to stop the main traffic constantly to let "nobody" to pass. The only reason why it's not implemented in crosses of dense urban areas is because it would go out of sync with other crosses nearby. There used to be a system for your smartphone to suggest the speed to maintain to find the next crossroad green. But it had too few crosses compatible and especially you were still gonna face people not using it which would form a line that would not go away in time, rendering the whole system useless.
I fully get the premise, but it could have been backed up by telling us when you think this bubble will burst? It seems more you suggesting that AI will trickle away from the bull market rather than implode.
I'm pro AI but I'm the first guy to admit it's a bubble. I just see it in the same way as the railway bubble or the internet bubble, as you need someone to do the grunt work of building these thousands of miles of track, put routers in place, or in this case train these giant language/diffusion/what have you models. It's not like we didn't end up using the internet, it just took wee while for people to come up with viable uses for it.
AI will be brilliant when the dust settles in speeding up and automating things, we've already got some coding assistance in programming which can be useful (mainly for typing out simple, mundane things). But right now it's all overhyped nonsense.
Hi Sasha! Nice video as always! I would like to hear your thoughts on the new end to end FSD software from Tesla. How do you think this will impact the valuation of the company and what do you think will be a realistic timeline for the deployment of FSD to everyone and the deployment of Robotaxi’s. Do you think this is a hype as well or do you think this will be a real thing
Good points. I'm reminded of when laptops maxed out their potential in about 2008 or 2009, when there really wasn't anywhere else to go with silicon chips.
I think people are 1. Noticing the shortcomings of "AI", and 2. Realising the use cases of "AI" are a lot more limited than people were hoping it would be. I'm just waiting to see how long it takes the masses to realise these facts, and how long until the bubble bursts.
After studying Machine Learning and then neuroscience, the current panorama is really curious. There is a lot of hype directed in the wrong direction, and a lot of emotional reaction to that hype is just as erratic. This video is an example of the latter. Unfortunately, understanding the current changes and development of AI architectures is tedious and requires daily effort (unlike previous technologies with which AI is attempted to be compared in a negative way). It is not something that is going to be understood by discussing "tech bros" and NFTs with an angry face.
@@SashaYanshin The premise of your video is that tech bros swapped NFTs for AI, creating the new bubble. Let's not be hypocrites. The problem with that is that the bias is obvious, and points that do not contribute to confirmation bias begin to be omitted. For example, saying that there is a long way to go before we can make a movie, when in practice three years ago people would see it as science fiction to have something like KLING or Gen 3. And this is due to the continued underestimation of human intellect and the potential of new architectures. Without going any further, GPT4o mini came out yesterday, and demonstrates an effort in optimization, with a better model than the original GPT-4 but extremely cheaper and faster. For "some reason" it does not occur to you that such developments in architectures can also be adapted in due time to other generative AIs. In fact, Gen 3 makes videos far superior to a year ago in under a minute. I'm just giving an example of how obvious confirmation biases lead to talking about silly bugs like Google recommending eating rocks and omitting anything that clearly hints at feasible rapid development in the short term.
@@cesar4729 objectively investors have swapped NFTs for AI, this isn't even up for contest, you probably need to ask ChatGPT what does hypocrite means too.
Yep, AI is a scam. The only difference between Block chain and this is when the bubble bursts we might actually have uses for this. Also I strongly disagree with your example for junior developers - I would never tell a junior to use AI, it's too inaccurate and will teach them to rely on it to code instead of learning which will make them awful programmers, there's even been a small study which finds this to be the case. I won't be surprised if in a decade or so we suddenly have a shortage of competent developers because we kept telling people AI will replace programming.
the fact that you use the word AI and calling it a scam shows you know nouthing about AI.. there are use cases for machine learning and deep learning that fall under the AI umbrella.. it works.. I use ML/DL everday for pattern recognition and removing the noise from the real data without damaging the real data which humans cannot as easily separate out with manually tuned signal processing tools without damaging the subtle real data.. Is "the all encompassing AI umbrella" as grandiose as its put out to be.. certainly not.. but AI is not a scam..
@@Mellowyellow8888 if you read my comment you would of realised I said there are actual use cases for this. But the fact is a large part of the hype is based on scam products or products that exaggerate how good it is.
AI is not a "scam" in the slightest. It's the next stage of our technological development, whether you like this current form of it or not. And it has nothing to do with "block chain", quit trying to connect the two, that's a false narrative the anti ai art community came up with. In a decade, people who want to be developers, will do so to work on their own projects. AI is going to force us to implement major economic changes. I bet most of the issues you have with AI, are actually issues with how Capitalism uses AI. But the two are not compatible, and only 1 can be changed, as it has many times in the past.
I work in big tech and part of it involves building GenAI applications for customers. I'm personally not impressed by it. Customers often have sky high expectations of what it can do, but reality hits different
AI is not a bubble, this is the real deal, this will be the most explosive tech of all time, and the world will never be the same, it's coming for everything. Everyone even the leaders in the field under estimate its progress and value, the unique thing about ai is it makes itself smarter, exponentially....
Based on what data? LLM improve by putting high quality data in it. No high quality data left != no improvement. LLMs only improved so much because they had a lot of air up. Not anymore.
Exactly! Just like how British cow farmers made their cows healthier by feeding them left-over cow! Putting into them what they got out for exponential growth.
@FlorianUniversität Nope, still not proven wrong, even with the new version of chatgpt. LLM fundamentally rely on massive amounts of training data. If they stop feeding it with the training data, they are only able to "fine-tune" their models such that they look better on paper on their limited metrics. LLM are still not intelligent and still cannot get additional knowledge on their own. Also, only god and openAI know how much additional training data was fed into the newest version.
AI does not think... it is limited by what we already know. It is just faster at providing data, but in all reality it is just drawing on data from a huge database of scraped information. IMO, AI is a long ways off from thinking beyond what humans can do now. I've used many of the tools out there and the results are not awe inspiring... maybe the creation of images or video, but data is still data and it does not do data well, especially in relation to the task we do in the physical world.
Another point: Now that so-called "AI" has been fed (by humans) the content of the internet and starts making up things when it has no answer and puts those answers back on the internet, it will degrade since feeding on its own garbage can only cause more garbage to come out in the future.
This my 3rd AI Summer-Winter Cycle since I became a computer profession. Will be glad when Winter comes and we get back to good AI research and development.
@@SashaYanshin yes, what I mean is that the hype sometimes is justified. The problem lies on too much investment in a short timeframe (the bubble). When people start losing money many would think that the tech is not important or valuable enough. In the case of AI I believe all the hype is justified but at the same time, the ammount of money poured into it might not give the fast return some expect, or in some cases many investments can even lose money. Still the technology is extremely amazing and will probably have very impactful results in the long term. I think I'm not in desagreement with the idea of the AI market bubble. Just wanted to say that even considering that, I think AI is a very important technology and that I'm still hyped for it. Some people compare it to Cryptobros NFTs hype, but it's a very unfair comparison, one is a useless market speculation trick, another a real world changing tech.
How if they don't have the GPUs???? Nvidia's GPUs are bleeding edge and double every year or so in performance. If DARPA had an AI 30 years ahead of our time, they'd need the compute of 2055, and that is not true at all.
Worth mentioning that Apple has prioritised running smaller models locally on device. They started kitting out Mac's with NPUs in 2020 and iPhones have been running tonnes of NNs for years. The companies training huge LLMs are all using GPUs, GPUs are amazing but not designed for training or inference. The situation changes when we get specialised hardware both sides of LLMs. Would be good to hear your thoughts if training and inference costs plummet.
Yeah google search is borderline unusable now. duckduckgo search is...OK I guess. At least it doesn't have the relentless product grifting and AI hallucinations!
No idea what you're talking about. I have never talked about a house price crash. There is a very real mortgage affordability problem happening right now - just look at housing costs and mortgage delinquency data. But that's a somewhat separate issue.
@@RabJ208 I made a video reporting on data coming out at the time in multiple reports. Then spent the rest discussing mortgage affordability issues. I suggest you actually watch my content over time before taking issue with points I made or didn't make.
@@SashaYanshin, you said: "House prices are beginning to crash and it's not surprising due to vast increasing mortgage intrest rates" - Do you still think higher intrest rates correlate with lower property prices? If so, why are property prices rising? Take NI for example - we're up 5% year on year. Not surprising?
"AI" is actually algorithmic learning, nothing more. Its literally the most expensive, and energy consuming, google search. We are still over 10 years away from AI.
@@budderk1305 to be honest, I was one of those idiots. When I first used "AI" it was incredible. But after using it daily, learning the fundamentals, and going deep into its flaws, I realized that I've been lied to. Don't get me wrong, I found ways for AO to simplify my life. But it's not a groundbreaking improvement. The hallucinations kill the effectiveness of this software. And there's not a solution to correct that problem.
I work with a cloud based service that's pushing AI super hard. Every client wants it, they don't know what it is, what it does... I just recently participated on a call to discuss predictive insights based on existing data they have collected... It was straight up 1 hour dull of buzzwords and crap. Also the CEO has no idea what data they have or the state of it, his "goals" were basically :" magic"
it's refreshing to hear your opinion. But I'll say that as a developer gen AI has created a lot of value for me increasing my productivity. Overhyped? yeah, but it's not completely useless. Neither extreme is correct IMO
I didn't say it is completely useless. In fact I said the opposite - I use it every day. Just because it is useful doesn't negate the obvious fact that a huge bubble has formed.
He didn't say it was completely useless. The fact you are saying ''it's not quite completely useless' is testament to the shitstorm coming to Ai businesses after literally hundreds of billions of investment predicated on it being a complete paradigm shift not just to business but to our very existence. Reality is it's not completely useless. That is funny
Yeah, I think it's a useful tool but it really should have stopped there and have been used as a tool for workers rather than a replacement. I personally think Sasha is correct and too much money has been poured into it. AI is great at helping doctors diagnose, app designers/developers create and content creators understand trends but it's not going to replace them completely. Humans are currently the ones that have to still rectify AIs mistakes and understand when AI has started to harvest data from incorrect sources. My other issue with 'AI' (more of a personal one, unrelated to the video) is how it's exponentially increasing the amount of spam content which is especially easy to initially overlook now that short form content is so popular. I hate the 'AI' spam!! Very interesting topic though. Good video and interesting comments. :)
I believe the AI Hype bubble is gonna burst. Not AI itself. I think alot of people got too hyped on it and now it's being exposed as a tool and less so a means to replace you. I know people are saying give it 5 - 10 years then you will be eating your words but right now it's just a hyped up tool.
at first I was shaking my head in disagreement. Listened attentively nonetheless, couldn't navigate away somehow... and I loved this video. Well done. You've won me over,.
NFTs and Crypto Currency never impressed me, but AI has already shown it has great potential. Sure, not all companies will be successful and many will fail, but AI is here to stay
@@piccalillipit9211 he said it has potential, like, look how generative AI couldnt do hands and now it can, or how it can simulate a human voice that hears so so real with laughs and breathe pauses, etc (chat gpt-o), how we could never imagine to generate a video from text and then it does, etc... none of it its perfect but it has a lot of potential if it reaches a nice level of quality
@@gatolocoowo Yeah - great. What can it do? what use case does it have? Cos hallucinations are a FEATURE that will never ever go away. So that means it cant be used in any regulated industry. It cant fly a plane cos X% of the time "crash into the mountain" will be a legitimate action, it cant be used in finance cos "lose your client money" is a legit probability - it has not idea of good / bad or right and wrong. It is NOT Generative Artificial intelligence at all - Its a Generative Probability Engine. so it can only be used in applications where "machine does bad thing" is acceptable. Cos there are an infinite number of bad things, and it WILL do them ALL use cases are predicated on it NOT being what it is.
LLMs and other ML models? Sure. They're here to stay and I can say that confidently because they're frankly nothing new. In the hands of trained professionals they can be game changers, especially in science. But not everything needs an "AI" chatbot to give you confidently wrong answers - especially when it costs an arm and a leg (and a fuckton of carbon emissions) to run.
technologies can be incorporated just as much as they can fall out of favor. If in the end the costs will be to great to possible profit - AGI is a gambit because we know it to be a sci-fi dream - we'll see for how long it is here to stay
Ai would fucking hate my job more than I do so I’m not worried.
🤣
❤
I'm not worried until they make a robot that can run a sewage treatment works including cleaning up gallons of shit when things go wrong lol
It can automate the shit outta your job and pass it other sub processes to run the random shite. It sees your hate as the problem to solve.
P.s: thats the ideal AI anyway, probably take us 5 years or more
@@boyasaka Sewage workers are the most valuable people in society measured by number of lives saved per day worked...!!!
My company is producing "AI software" for the NHS. Its not AI. Its an automation algorithm using SQL that could have been done a decade ago if people coulr have been bothered. But hey, its AI now!!! Buy our AI!!!
Ripping off the NHS is a long-standing tradition I am sorry, as a tax payer, to say.
I'll take two.
@@theastuteangler 😂
Sounds accurate.
At least you used a database, Amazon's AI shopping used a bunch of people in India as the backend. LOL!
All I want ai to do is when I look at a grainy cctv image I can say 'enhance ' and it zoom in for a perfect crystal clear image. Just like the movies.
This is already somewhat possible
@@TheRealHassan789 "somewhat" being the key word.
@@TheRealHassan789 It's not, the end result is a bit shit. And this is basic stuff
@@TheRealHassan789 You cannot make detail out of nothing. AI will do it, but it will just be a hallucination, which is absolutely pointless. Actually, it's a great metaphor for how useless AI is.
@@d0m96 check this paper out: “Personalized Restoration via Dual-Pivot Tuning” … it’s pretty cool. Because it lets you bias the “noise” with known facts. Hence AI can actually do a pretty good job.
So when they need the content the social contract is that it's free use, but when it comes to torrenting tv shows and movies it's like stealing a car? 😂
Exactly.
Also ignoring that free use implies that the content is being used in either a transformative manner or for education purposes. Using it as training data for a commercial product should not count, as the content is not transformed in any way by the actual people handling it, or selling the product it was used to train. In fact, only the end user should be protected, as they are then using the data transformatively to generate things.
This
Because if everyone downloads movies for free then they will be financed how exactly? How will all the people working on them pay their rent?
@@robertruffo2134they can go ahead and eat the bugs
I work in a corporate office for an international company. As an IT support manager, all my leadership keeps asking "cant we automate this with AI" ... they say this about everything. My plan is to stay human, analog and manual. In the future, that will be the specialty..
“There’s no place like home.” We had the power all along.
"AI could help with automation, but also need developers. So money please"
Those same people two years ago, when asked if there was a budget to create traditional automation tools (which most of these "AI" products are anyway), would claim it wasn't worth it.
tell them to use VBA in Excel 🤣
How many people in your office have Mac addresses?
7:56 that Microsoft AI CEO is so full of sh*t. Everything uploaded on the internet is freeware? What planet is he living on?
The amazing thing is that all the tech bros and young keyboard warriors support this sort of statement because they completely do not understand the implications.
@@SashaYanshin so Microsoft is ours now as well
Everything is free for the taking regardless of laws, ethics and common sense but don't you DARE take away from them or they will run after you. That's the average way of thinking of these people.
He also implied Windows OS and Office suits are freeware. I love that. 🙂
Unless it's Microsoft's sh*t of course, that's copyrighted ... also reason why AI is cr*p ... internet is full of cr*p, so cr*p in, cr*p out ...
Too many companies are claiming to be AI when they are not . Its become a buzzword to be accepted by places with nothing more than a normal software algorithm.
Very accurate. And this is exactly what a bubble looks like every time too.
🤣
lol. I guess thats what algorithms are: AI.
Ok, which company actually has “AI”? I mean, real AI, not just a pre-trained neural network.
@@andrzejostrowski5579 There are no real AI at this time. All these fake "ai" is just algorithms, more powerful, but still it's just algorithms. 😂🎉
And from a creative point of view, it’s hilarious how much effort, and resources are being pumped into tech that gives you less control
A friend wants to remove 90% of the devs from his office because AI can now generate code. He wants to keep a few to manage the code. What he doesn't realise yet, is that although useful for some code ideas and snippets, a lot of what it churns out is bollocks and you have to correct the AI and make it do it again. Then you have to figure out how to add that code into your own. We are far away from replacing humans in some roles, but people really think that AI is better than it is right now.
I find it useful as a better version of smart complete; but you still need to know what you want it to do, how to architect a larger system and proof read what ever it spits out.
I used to work for a company making packaging machinery, our customers would always insist that we should have the latest controllers, use the latest motor technology and have lovely HMI's everywhere, all of our competitors did.... this was all nice and good, we could do all of these things but at an upcost to the customer. I explained to many customers that our standard machine offerings with their basic control systems were proven, worked well, were easy to operate and maintain and cheaper than adding the technology they wanted, our standard machines also had a much quicker turn round and we made more money on them. The question I ended up asking our management was do, we really want all this technology on our machines if it's not really that beneficial, is more complex and is costing us more? The answer was that we need to use this technology because all of our competitors do, we can't be left behind... perhaps that's the issue with AI we have it because we have to.
That is part of the problem, yes. Good example. You can see this happening already with smartphones.
Asking Sora to tweak a video is like asking a painter to paint a new version of a masterpiece without looking at the original, with a few "changes"
Calling these LLMs "AI" is a stretch imo. I am surprised that this hype even caught on since these were a thing for quite some time, and were never ever envisioned to be extremely groundbreaking and paradigm shifting ever. At least nowhere near the potential the advent of widespread internet coverage had.
It is ridiculous how many promises are made regarding AI generating text, music, code, pics videos and ultimately be smarter than humans. And all it is is a glorified optimization algorithm that finds a most likely word, note, token, picture or video fragment that would fit the best given the input requirements and limitations. Sure it will be able to drive cars anywhere in the world _someday_. Or deliver pizza. Or perform surgery. But the current AI will never be able to invent.
i recently did an interview with a startup company in germany that is using ai to attempt to automate trucking. the job is working at a desk with a gaming wheel and watching 4 trucks do their work at a port, when the truck gets stuck you take over and help them out. meaning a single driver is technically driving 4 trucks at once. i think most ai will end up in a similar capacity, reducing the need for labour. otherwise the entire planet will grind to a halt because a bird flew in front of a camera.
😂😂 Western companies have already failed on automatic 😂
So, you could rob a truck without the need for hurting anyone? And you could strip them for parts too?
Cyberpunk esque call the police when danger is detected sensors, or other defences could be easily circumvented with the right knowledge. So robbing trucks
What happens if TWO trucks get stuck at the same time?
What happens if three or 4 get stuck?
Is the port going to allow the company time to play with their legos, and stall the whole operation?
How many truckers salaries do they think a full hour of a whole port not operating is woth?
So if the person falls asleep at the wheel there will be four crashes not just one lol more damage per point of failure, love it!
I agree with the points raised here. If this technology could be used to complement truck drivers, it would be fantastic. Imagine a situation where a driver falls asleep at the wheel, and the system triggers an alert, allowing a virtual assistant-enhanced by advanced tech but still human-led-to intervene and prevent a potential tragedy. However, the real issue is that some companies are more focused on cutting labor costs and maximizing profits rather than enhancing safety and support.
Why isnt ai replacing all ceos and company presidents? Those folks all need to be on the unemployment line.
They'll be able to make more money and do even less work.
I knew the bubble had burst when I saw that you can buy AI powered shower gel.
Mine was a box of an AI "power toothbrush" at the supermarket. There I understood how much of a farce this is becoming.
Fridge, and apparently some saw a rice-cooker
My samsung washing nachine has 'AI'. All that AI does is just recall the most used laundry cycle upon startup
I will not pay for artificial intelligence when I can use real intelligence for free.
I use both
Most funny boomer
Pretty sure that humans require payment too
And the funniest is there is no such thing as AI at the moment:DDD
the irony is they make no money either way - if the hype was real (its not) and they put everyone out of work - whos paying for ther subscription ?
B2b exists
@@TheManinBlack9054 Sure but business's dont exist with out paying customers who is apple selling phones too ? or google showing adverts too ?
@@TheManinBlack9054You’re forgetting about the consumer. Christ, who’s buying the goods if AI has replaced said people in those jobs. Ten points for you being an absolute genius
No, that's actually the most terrifying aspect about it. If nobody has any money and become dependent on "UBI" to live, we will devolve into a dictatorship rather quickly, as the goverment (and their friends) will have 100% control over your life. I think that is the whole point actually, but that is another conversation.
@@IsCaRzY1 I asked this same question 2 years ago nobody had an answer It seems like everyone is catching up😂😂
I'm a lead in a team building AI and process automation solutions. You are 100% correct on this.
AI has found its natural level. Gareth Southgate parody videos.
Those are good!
It's clear the Gareth Southgate AI videos are of the highest quality order and have reached General AI. They could probably come up with a better strategy than the real man.
AI is making the calls on England
@@kippsguitar6539 I think Ai took over in the last game against Holland. Southgate has been binned l.
The problem with a lot of A.I. as I see it now is the amount of time, money and effort that is being put into replacing people and doing general tasks like video creation, chat bots and investment tools. The system is logic-based and should be geared towards managing and improving things like urban planning, resource management and advanced robotics.
It's easy to excuse cases where AI messes up generating things like images and videos but giving AI complex tasks could cause catastrophic results if it messes up.
@@rayauxey I get where you are coming from but that just proves further that it isn't yet fit for purpose if people are afraid to, or unable to use it where it would be of most benefit !!
The tasks that "AI" would be good at is managing efficient workflows. But wait...thats Business Process Automation (BPA) and that's been around for many years.
Conventional ML techniques work better for those kinds of things as they're orders of magnitude more efficient and accurate. LLMs are mostly a very compute heavy way to generate text (or images).
Those things don’t increase $ and that’s the main focus in America
Hold S&P through highs and lows. Invest regularly. Then all this noise just becomes entertainment instead of stress.
Nailed it. 🎯 "croac-croac" 🐸
Absolutely
I agree generally but S&P 500 is very top heavy with these handful of AI hype companies like MS, Meta, etc. diluting the diversity appeal of a S&P 500 fund. They do have solid underlying business, but they will take a hit along with S&P 500 funds when AI bubble pops.
Look at the top 7 holdings of SPY and you'll see that they're all AI related plays. (i.e. MSFT, GOOGL, GOOG, META, AAPL, NVDA, AMZN). Collectively they make up 31+%. And that doesn't include other names that may have been pumped up by the AI hype also in the index (i.e. smaller semi companies, data center reits, etc.). In other words, thinking you're diversified holding a S&P 500 index fund and won't suffer from the fallout of an AI crash is fantasy. S&P500 has basically become somewhat of a AI theme momentum fund.
Where have you been? I have missed your perspective and your sarcasm you always make me laugh❤
Google AI told me that a conversion from 107kg to Stones and Pounds was over 31stone (It's actually 16st 12lbs). It even laid out the steps as to how it got to the answer. Problem there was in the explanation it arrived at a figure of 8st 12lbs! But still told me it was over 31st. I've also had to get rid of Google Gemini and revert to Google assistant as it just kept getting answers wrong.
Doesn't standard Google search already have a unit converter?
@@carultch Yes, but they put their AI (Gemini) on Android phones as default, to replace Google Assistant. After this incident however, I have reverted to Assistant.
@@richard127gmso they put a new intern in the place of the employee, and tried to convince people they were the new savont
*IT APPEARS WE HAVE REACHED PEAK TECH*
as someone said "“we perfected the internal combustion engine in about 1940 - since then it has just been minor incremental improvements - 80 years of tiny advancements - maybe we have reached peak tech"
Atmospheric engine was peak steam until James Watt came along.
But at least cars became reliable ...
@@adamaglionby7438 Actually James Watt's engine was an atmospheric engine but with incremental changes to Newcomen's invention, yielding impressive gains in efficiency. The first true steam invention was I believe down to Trevithick, in about 1801. I have a soft spot for Trevithick: he won a bet on his steam-driven "road" engine ("Puffing Devil") being able to move coal near Merthyr Tydfil better than pit ponies. But he was so happy to have won that he retired to the pub with his mates, leaving the engine puffing away, as a result of which it then blew up.
In 1698, Thomas Davey sold commercial steam pumps.
And of course the Aeolipile.
@@mikerodent3164 Then we invented the steam turbine - which we still use in nuclear subs and carriers and power stations
This is why Microsoft now want access to everything you do on your computer. The low hanging fruit on the internet is gone so now they will scrape your every minute of life.
I hope your correct and there are no emergent properties from this investment bubble, because with bugs it would be dangerous if it did continue to be more advanced while profit driven, for something all controlling and powerful there can be no bug ever or its all over. Luckily there is SSI with Ilya Sutskever, machines creating better machines in a closed loop are a powerful thing. But with no product ever, only one final release, it's investors are investing in 100% hype.
The AI in Google search isn't useful at the moment, but it is learning from everyone that clicks on an AI result. We are teaching it, correcting it, improving it. At the moment it says 'oh, you are quite correct to point out my error. I am still new and may not get everything right just yet'. In a year it will say 'Sorry, but my answer is correct - you are in error. Please wait while I change all references on the internet and erase all your personal information according to the Computer Heretic act of 2026.
This is another AI grift. Companies do use user interaction and feedback as signals to train and optimize their models, but it's happening with A LOT of engineering effort, and it's just a small part of the overall work. It's not even close to "the AI is learning from you", nowhere close.
Such day, older books will suddenly be worth their weight in gold.
Google doesn't give a shit if the answers it gives are incorrect. More inquiries = good for business, even if it means their service is poorer in quality than it was before. All they care about is how much time you stick around taking in ads.
AI chatbots are where a lazy human asks a lazy robot for answers. The answer is usually wrong or incomplete, so the human spends an hour chatting to it, trying to make it understand where it went wrong and apologise.
Sadly, typing "understand where you went wrong and apologise" does the same thing as that hour of chat It still won't understand, and it isn't really sorry - but it will give you another dubious answer.
Ai is already smarter than half the human race. Soon all the human race. You aren’t a smart human if you don’t utilise AI
Grandpa, those are llms not Chat bots and the answers are not usually incomplete or wrong.
ChatGPT is extremely useful as an LLM.... I use it all the time now in tech.
Garbage in garbage out
Is still undefeated
You make a good point about the sources required for AI to continue to make gains. This is probably why Adobe tried to muk with their terms of service to allow them to utilize their customers photos to train their model(s).
Adobe’s AI generation is pretty horrible too. Probably the only useful improvement is generative fill for removing elements in an image or extending edges.
Even if the scaling laws are correct, nobody is going to be convinced by tech bros saying "please gimme 1 trillion to build 3 nuclear power plants and a teradata center, the result is guaranteed AGI!!!" The problem is, we need those trillions to do other things now, like heating, air conditioning, building houses, all that boring stuff.
On Point Brother... on Point
Thank you!
Ah the smartphone analogy - bought one in 2020 for 200, now a new one of the same type is selling for 250 and is still competitive, the only upgrade I consider is a new memory card that doubled or quadrupled in the meantime
Wait, all content on the internet is free use and that has been generally agreed upon since the 90s? And here I have been diligently paying for my movies, games and software for years...like a fool. That CEO was literally saying that piracy is ok.
Reasonable point of view. I think the problem with AI right now is that the whole market is inflated by companies that don't generate actual revenue with the tech (at the moment). Everyone is investing billions but these investments don't translate into actual useful products that can be sold. But I'm not sure it's as obvious as you said because I have no idea how long it will take to create functioning products. Do you think that things like tesla robo taxi aren't realistic in the near term?
As soon as I saw a robot in a sushi place carry drinks to a table, I realized we had that same thing since 1985. 😁
Correction: GENERATIVE AI has become bubble. Regular AI is doing fine - finding cancers, diagnosig patients, doing optimisations and defectoscopy....3:51 - same about Ukrainian War or any other subject realy....
I work in IT leadership role and I have an innovation team looking at AI more specifically LLM and I can’t find a single commercially viable use case other than chat bot even then it’s not that commercial. I also recently completed an AI course at Stanford and looking at the latest in AI it’s actually very far away for anything to be fully automated. We’re talking another 10 years at minimum.
Yea advertising, entertainment, automation, heatlcare, Agriculture, weather forecast and so on but you only think about chat bots. You're the average IT boss, useless and years behind.
@@Torbintime rather than throwing mud why don't you share some tangible commercially viable examples then?
@@drtomgao "We estimate that generative AI could increase labor productivity by 0.1% to 0.6% annually through 2040. Combined with other technologies, work automation could add an extra 0.2 to 3.3 percentage points to productivity growth." says McKinsey. I personally think ai assistance in costumer support will absolutely take over, it's a way better user experience than having to wait 10 minute just to talk to some Indian guy. Entertainment is gonna be crazy, there's already weirdos who date some ai now imagine if there was a decent actual ai attached to a robot, millions of people would buy that.
@@Torbintime ok so based on your comment you're just making assumption based other people's view and comments. You haven't received direct education and looked at the maths behind LLM nor understand cost of computation of inference/prompts nor understand costs of IT/AI projects and governance of AI hallucination risks. Lots of unfounded assumptions not based on a single tangible example of a commercially viable usecase in 2024. I can also quote lots of benefit for AI in 2070 but we're talking about right now and the next 5 years.
@@Torbintime those are very modest expectations for a bubble valuated in the order of HUNDREDS of BILLIONS of dollars. AI being a bubble does not mean that AI itself is useless, just that the expectation is vastly out of all proportion. with that kind of money, if AI fails to replace literally all white collar jobs by, uh, next year, it's going to fail catastrophically. literally OpenAI and Anthropic are running out of outlets and are starting to scrape the bottom of the barrel, starting with the saudis. When you see sam altman begging softbank for money, you better cash out.
I am 55 years old and remember the last time AI was over-hyped and then AI Winter happened. I am currently investing in a money market fund.
Which?
@@daveabc12 can't answer for him but I took CSH2 ETF for my investment accounts and switched 25% of my pension pot into their in house MM funds.
@@daveabc12 Scottish Widows Active Money Market fund. It's the only one available in my company pension. This is where new contributions are going. Previously invested money is staying in passive global equity index tracker and over five year inflation linked Gilt tracker funds.
@@daveabc12 The only money market fund offered in my company pension options.
Often a good idea isn't self sufficient, but dependent on other parameters.
Internet, computing power and the cost of computing power: when that matured, LLM's became opportunistic, and it's a magic wand.
We shouldn't be angry at a dog who doesn't pronounce all English words correctly.
thank you for this video
🙏
i have no idea why this bubble wasn't even created when the tech was already being developed
now they just rebranded the same old machine learning and neural networks as AI and now everyone is investing
It's LLMs and text2image and recently you see it more accessible easy to use and in the hands of the general public mainly but the typical consumer/investor doesn't know that, they only know the general "AI" buzzword which is all that really matters from a marketing standpoint
Because major part of stock ownership are USA citizens, and in USA every second MacDonald's employee buys stocks. An investor mob like this is a blind trend follower. USA top money people did a good job of getting this crowd onboard, much easier to fleece compared to specialized investor.
Saturn, is there life on Titan?
Superbly oratorically articulated and evaluated. Thank you Sasha👍👍👍
Thank you so much!
Finally someone with common sense. Tell me one thing I can use the AI for in normal life. Nothing and they keep pumping billions into it.
Autonomous cars; Robots 😊
AI is likely overhyped but the fact that you don’t use it is hardly a strong argument against it.
5 - 10 years from now whenever you call or send a chat/text message to your bank, phone company, utility company etc. you will talk to AI instead of a person. If you complain enough and ask enough times it will transfer you to a real human, but most likely it will answer your questions better and faster than a massive call center of people who don’t really want to be there can or ever will. This transition is already in progress. Advances in robotics opens up manufacturing, construction, farming, food prep, transportation, repair and maintenance work to AI as well. Basically every blue collar job will eventually be done or assisted by robots running on AI. AI and robots will help us get through the population collapse which is currently in progress all across the developed world. It will allow us to do more with fewer people. It won’t result in no one having a job. People will still have jobs, but they’ll be different than they are now.
@@EamonnMurphy-cc1dt autonomous cars are not AI same as all other robots just preprogrammed following specific instruction sets based on "IF this happens DO that"
@@Cfb2987 yeah right keep dreaming
Often, we have the truth right in front of us and either fail to see it or choose to ignore it. Thank you for your efforts in spreading this important truth about the AI craze.
Great content as always. When do you think it might happen?
Sasha is absolutely correct. This is a massive bubble. Curent AI is just a fancy data fitting method. What will be truly revolutionary is AGI. And we are as far from AGI as mammoths from social networks. I am not even sure if AGI is possible. It would take a completely new approach and understanding of consciousness and stuff. AI has peaked. AI will not get significantly better than it is now for at least hundreds of years.
This 100%! One thing that makes humans human, is this collective socially constructed lie that symbols (words, pictures, sounds, etc.) mean something and aren’t just arbitrary connections. AI simply can’t lie. It can tell falsehoods (purposely programmed or by accident), but it can’t inherently live in our social lies.
AGI is also not a magic bullet. Creating AGI simply means creating another sentient being. What's the guarantee that AGI will even want to do chores for humans
Very timely and spot on, SUBSCRIBED!
Ok I laughed so freakin hard at that AI response to the question "how many rocks should i eat per say". Thank you for that Sasha, after watching the debates I seriously needed a good laugh.
Same! Trillion dollars well spent 😆
The debates weren't enough of a laugh?
Always a voice of reason
🙏
AI is just big data rebranded as AI. Sh×t goes in, sh×t comes out
it's like the underpants gnomes in South Park -- Step 1: collect underpants, Step 2: ???, Step 3: PROFIT!
Nonsense can big data, Draw pictures, write poems, translate between languages, speak naturally, understand humans, code, make music, operate robots...
@@TheReferrer72 yes, it can.
@@andrzejostrowski5579 no it cannot. there are many technologies happening there
Just like "Cloud" is just server client. We've had that since at least the 70's.
What a great video. I find it funny that all of a sudden, all these companies have AI. How did they get to it? It is amazing. However, if you use Google Maps and use to get directions, the thing is so bad, that at this second, I am on I4 and if I am in traffic and 508 passes over me, I am magically on 508 and get a new route on how to get to my destination. Unbelievable. Or, that Google shows me a traffic jam ahead because there are a bunch of cyclists standing on the side, or that it gives me a route through a private property, or a route through a dirt road if there clearly is a better road, or 2 other suggested routes that make no sense at all. I always wonder how these valuations come to live if companies are clearly so bad in day to day things and with this AI stuff, it gets even worse.
Less than More 10 trillion dollar valuation. Without making single dollar in net Income. I guess we don't know better. Certainly smarter people than us are r this show.
Just one more AI forcefully invading other people's cozy space and hobbies, bro. It will make everybody a trilionaires this time. /s
THANK YOU
I would love to see AI working at traffic lights so when I’m sitting there at a red light watching a green light with absolutely no bloody cars wouldn’t happen.
Until it turn into the AI from Eagle eye movie !
We've had camera pickups in traffic lights for over two decades. Whether they're widely deployed anymore, I don't know. My father used to install them for local government.
Why do you need AI for that? AI implies ability to learn from new data and change its behaviour. All you need is a normal script with presence sensors to detect vehicles waiting in a lane but not in the other. Such system is already implemented in some places since years, although rarely. Usually it's crosses between very tiny roads with negligible traffic crossing important veins that do not need to stop the main traffic constantly to let "nobody" to pass.
The only reason why it's not implemented in crosses of dense urban areas is because it would go out of sync with other crosses nearby.
There used to be a system for your smartphone to suggest the speed to maintain to find the next crossroad green. But it had too few crosses compatible and especially you were still gonna face people not using it which would form a line that would not go away in time, rendering the whole system useless.
I like how we learned nothing from Crypto and NFTs, just jumped on the next thing.
Sasha is a boss. Always calls out the BS and keeps the recipts. Thank you for keeping me sane during the last big dip hahah
Will keep saying things as I see them! Thank you!
One of your best videos, a factual AI journey. A multiple watch.
∆p --> 0 as $ --> ∞
∆p being change in performance of AI
You heretic! this is the opposite of exponential growth! How dare you?!
@@SashaYanshin I KNOW - line goes horizontal not line go up
Off subject, do you have any opinions regarding shorting the banks holding commercial real estate?
I fully get the premise, but it could have been backed up by telling us when you think this bubble will burst? It seems more you suggesting that AI will trickle away from the bull market rather than implode.
Great commentary and delivery!
I'm pro AI but I'm the first guy to admit it's a bubble. I just see it in the same way as the railway bubble or the internet bubble, as you need someone to do the grunt work of building these thousands of miles of track, put routers in place, or in this case train these giant language/diffusion/what have you models. It's not like we didn't end up using the internet, it just took wee while for people to come up with viable uses for it.
AI will be brilliant when the dust settles in speeding up and automating things, we've already got some coding assistance in programming which can be useful (mainly for typing out simple, mundane things). But right now it's all overhyped nonsense.
Hi Sasha! Nice video as always! I would like to hear your thoughts on the new end to end FSD software from Tesla. How do you think this will impact the valuation of the company and what do you think will be a realistic timeline for the deployment of FSD to everyone and the deployment of Robotaxi’s. Do you think this is a hype as well or do you think this will be a real thing
good one, subscribed
🙏
Good points. I'm reminded of when laptops maxed out their potential in about 2008 or 2009, when there really wasn't anywhere else to go with silicon chips.
I think people are 1. Noticing the shortcomings of "AI", and 2. Realising the use cases of "AI" are a lot more limited than people were hoping it would be.
I'm just waiting to see how long it takes the masses to realise these facts, and how long until the bubble bursts.
After studying Machine Learning and then neuroscience, the current panorama is really curious. There is a lot of hype directed in the wrong direction, and a lot of emotional reaction to that hype is just as erratic. This video is an example of the latter.
Unfortunately, understanding the current changes and development of AI architectures is tedious and requires daily effort (unlike previous technologies with which AI is attempted to be compared in a negative way). It is not something that is going to be understood by discussing "tech bros" and NFTs with an angry face.
Must be hard doing everyday things with a head that big!
@@SashaYanshin And what would that fifteen-year-old sarcasm be contributing to the dialogue?
@@cesar4729 Same as your insinuations and points about a few words in a 15-minute video
@@SashaYanshin The premise of your video is that tech bros swapped NFTs for AI, creating the new bubble. Let's not be hypocrites.
The problem with that is that the bias is obvious, and points that do not contribute to confirmation bias begin to be omitted. For example, saying that there is a long way to go before we can make a movie, when in practice three years ago people would see it as science fiction to have something like KLING or Gen 3.
And this is due to the continued underestimation of human intellect and the potential of new architectures. Without going any further, GPT4o mini came out yesterday, and demonstrates an effort in optimization, with a better model than the original GPT-4 but extremely cheaper and faster. For "some reason" it does not occur to you that such developments in architectures can also be adapted in due time to other generative AIs. In fact, Gen 3 makes videos far superior to a year ago in under a minute.
I'm just giving an example of how obvious confirmation biases lead to talking about silly bugs like Google recommending eating rocks and omitting anything that clearly hints at feasible rapid development in the short term.
@@cesar4729 objectively investors have swapped NFTs for AI, this isn't even up for contest, you probably need to ask ChatGPT what does hypocrite means too.
Yep, AI is a scam. The only difference between Block chain and this is when the bubble bursts we might actually have uses for this. Also I strongly disagree with your example for junior developers - I would never tell a junior to use AI, it's too inaccurate and will teach them to rely on it to code instead of learning which will make them awful programmers, there's even been a small study which finds this to be the case. I won't be surprised if in a decade or so we suddenly have a shortage of competent developers because we kept telling people AI will replace programming.
That’s not what I meant. I said the AI is like having a junior developer helping you. 👍
Blockchain is based on true great math, and it's a crystal clear techno. No matter how one uses it (for crypto hoax or proof of work)
the fact that you use the word AI and calling it a scam shows you know nouthing about AI.. there are use cases for machine learning and deep learning that fall under the AI umbrella.. it works.. I use ML/DL everday for pattern recognition and removing the noise from the real data without damaging the real data which humans cannot as easily separate out with manually tuned signal processing tools without damaging the subtle real data.. Is "the all encompassing AI umbrella" as grandiose as its put out to be.. certainly not.. but AI is not a scam..
@@Mellowyellow8888 if you read my comment you would of realised I said there are actual use cases for this. But the fact is a large part of the hype is based on scam products or products that exaggerate how good it is.
AI is not a "scam" in the slightest. It's the next stage of our technological development, whether you like this current form of it or not. And it has nothing to do with "block chain", quit trying to connect the two, that's a false narrative the anti ai art community came up with.
In a decade, people who want to be developers, will do so to work on their own projects. AI is going to force us to implement major economic changes. I bet most of the issues you have with AI, are actually issues with how Capitalism uses AI. But the two are not compatible, and only 1 can be changed, as it has many times in the past.
I work in big tech and part of it involves building GenAI applications for customers. I'm personally not impressed by it. Customers often have sky high expectations of what it can do, but reality hits different
AI is not a bubble, this is the real deal, this will be the most explosive tech of all time, and the world will never be the same, it's coming for everything. Everyone even the leaders in the field under estimate its progress and value, the unique thing about ai is it makes itself smarter, exponentially....
What is claimed without evidence can be dismissed without evidence.
Based on what data? LLM improve by putting high quality data in it. No high quality data left != no improvement. LLMs only improved so much because they had a lot of air up. Not anymore.
Exactly! Just like how British cow farmers made their cows healthier by feeding them left-over cow! Putting into them what they got out for exponential growth.
@@realkyunuAlready proven wrong 😂
@FlorianUniversität Nope, still not proven wrong, even with the new version of chatgpt. LLM fundamentally rely on massive amounts of training data. If they stop feeding it with the training data, they are only able to "fine-tune" their models such that they look better on paper on their limited metrics. LLM are still not intelligent and still cannot get additional knowledge on their own.
Also, only god and openAI know how much additional training data was fed into the newest version.
AI does not think... it is limited by what we already know. It is just faster at providing data, but in all reality it is just drawing on data from a huge database of scraped information. IMO, AI is a long ways off from thinking beyond what humans can do now. I've used many of the tools out there and the results are not awe inspiring... maybe the creation of images or video, but data is still data and it does not do data well, especially in relation to the task we do in the physical world.
You make some good points 🤔
🙏
Another point: Now that so-called "AI" has been fed (by humans) the content of the internet and starts making up things when it has no answer and puts those answers back on the internet, it will degrade since feeding on its own garbage can only cause more garbage to come out in the future.
Been waiting for this video for a while. Did not disappoint and I am only 6 mins in. 😂
Haha! Thank you!
This my 3rd AI Summer-Winter Cycle since I became a computer profession. Will be glad when Winter comes and we get back to good AI research and development.
The problem is not stocks of "AI" companies going up, the problem is to know when to buy, dump and/or short them.
Great content, thanks for making the general public aware of the imminent bubble disaster.
your spot on and soon people are gonna realise
I bought all these big-tech stocks 1.5-2 years ago. When do you think I should sell them?
Absolute mega video. Something I thought for a while, eloquently verbalised.
Thank you!
It had to be full with stupid ads!a functioning ad blocker,THAT would be a revolutionary invention.
heeeeeee's bacccccccccckkkkkkkkkkkkkkkkkkkkkkkkkkkkkk
2nd video already! :)))
the internet was a bubble. It doesnt mean the tech was false. Its normal for most new tech.
Yes… Exactly the same with AI.
@@SashaYanshin yes, what I mean is that the hype sometimes is justified. The problem lies on too much investment in a short timeframe (the bubble). When people start losing money many would think that the tech is not important or valuable enough. In the case of AI I believe all the hype is justified but at the same time, the ammount of money poured into it might not give the fast return some expect, or in some cases many investments can even lose money. Still the technology is extremely amazing and will probably have very impactful results in the long term.
I think I'm not in desagreement with the idea of the AI market bubble. Just wanted to say that even considering that, I think AI is a very important technology and that I'm still hyped for it. Some people compare it to Cryptobros NFTs hype, but it's a very unfair comparison, one is a useless market speculation trick, another a real world changing tech.
The irony is that DARPA’s AI is probably 30 years ahead of anything that has been allowed to be released to the public today.
DARPA annual budget is under $4 billion. It's not much in AI world. Besides they are more interested in AI that shoot then AI that talk ^^
How if they don't have the GPUs???? Nvidia's GPUs are bleeding edge and double every year or so in performance. If DARPA had an AI 30 years ahead of our time, they'd need the compute of 2055, and that is not true at all.
Worth mentioning that Apple has prioritised running smaller models locally on device. They started kitting out Mac's with NPUs in 2020 and iPhones have been running tonnes of NNs for years.
The companies training huge LLMs are all using GPUs, GPUs are amazing but not designed for training or inference. The situation changes when we get specialised hardware both sides of LLMs. Would be good to hear your thoughts if training and inference costs plummet.
Google search is now absolute crap.
Yeah - I keep surprising myself when I can't find the most obvious information that would be the easy #1 result 5 years ago.
Yeah google search is borderline unusable now. duckduckgo search is...OK I guess. At least it doesn't have the relentless product grifting and AI hallucinations!
I would like to know what search engine Google engineers use.
Hold AI stocks agi is now 3-5 years away I honestly belive this
And what about this big house price "crash" ?
No idea what you're talking about. I have never talked about a house price crash. There is a very real mortgage affordability problem happening right now - just look at housing costs and mortgage delinquency data. But that's a somewhat separate issue.
@@SashaYanshin, you made a video titled "house prices started to crash" when we were shouting from the roof tops that average prices were rising.
@@SashaYanshin, you've done a few "house price crash" uploads.
@@RabJ208 I made a video reporting on data coming out at the time in multiple reports. Then spent the rest discussing mortgage affordability issues. I suggest you actually watch my content over time before taking issue with points I made or didn't make.
@@SashaYanshin, you said: "House prices are beginning to crash and it's not surprising due to vast increasing mortgage intrest rates" -
Do you still think higher intrest rates correlate with lower property prices? If so, why are property prices rising? Take NI for example - we're up 5% year on year. Not surprising?
Good information, I didn’t know a lot of this. But now I do
Just FYI, the short position on Nvidia has grown from 29 million shares, to 315 million shares (10.8x increase) in just 14 days.
Yeah smart there definitely wasn’t just a 10:1 split 💀💀💀
Oh, finally someone is voicing my concerns about the AI hype out loud and bringing it to the masses. Thank you, Sasha!
"AI" is actually algorithmic learning, nothing more. Its literally the most expensive, and energy consuming, google search. We are still over 10 years away from AI.
10 years is probably quite generous...
@@adam7802The hype is that it will be here soon. There's no evidence that the progress is going as fast as the exponential investment growth.
preach and also lol at idiots buying into it
@@adam7802 maybe light years?
@@budderk1305 to be honest, I was one of those idiots. When I first used "AI" it was incredible. But after using it daily, learning the fundamentals, and going deep into its flaws, I realized that I've been lied to. Don't get me wrong, I found ways for AO to simplify my life. But it's not a groundbreaking improvement. The hallucinations kill the effectiveness of this software. And there's not a solution to correct that problem.
I work with a cloud based service that's pushing AI super hard.
Every client wants it, they don't know what it is, what it does...
I just recently participated on a call to discuss predictive insights based on existing data they have collected...
It was straight up 1 hour dull of buzzwords and crap.
Also the CEO has no idea what data they have or the state of it, his "goals" were basically :" magic"
it's refreshing to hear your opinion. But I'll say that as a developer gen AI has created a lot of value for me increasing my productivity. Overhyped? yeah, but it's not completely useless. Neither extreme is correct IMO
I didn't say it is completely useless. In fact I said the opposite - I use it every day. Just because it is useful doesn't negate the obvious fact that a huge bubble has formed.
nothing is utterly useless, even an xbox one s have some use to it
He didn't say it was completely useless. The fact you are saying ''it's not quite completely useless' is testament to the shitstorm coming to Ai businesses after literally hundreds of billions of investment predicated on it being a complete paradigm shift not just to business but to our very existence. Reality is it's not completely useless. That is funny
Yeah, I think it's a useful tool but it really should have stopped there and have been used as a tool for workers rather than a replacement. I personally think Sasha is correct and too much money has been poured into it. AI is great at helping doctors diagnose, app designers/developers create and content creators understand trends but it's not going to replace them completely. Humans are currently the ones that have to still rectify AIs mistakes and understand when AI has started to harvest data from incorrect sources. My other issue with 'AI' (more of a personal one, unrelated to the video) is how it's exponentially increasing the amount of spam content which is especially easy to initially overlook now that short form content is so popular. I hate the 'AI' spam!! Very interesting topic though. Good video and interesting comments. :)
@@SashaYanshin If you're using it yourself daily, then how is it a bubble?
I believe the AI Hype bubble is gonna burst. Not AI itself. I think alot of people got too hyped on it and now it's being exposed as a tool and less so a means to replace you. I know people are saying give it 5 - 10 years then you will be eating your words but right now it's just a hyped up tool.
😂😂😂 that ending killed me!
Thanks for watching all the way through!
at first I was shaking my head in disagreement. Listened attentively nonetheless, couldn't navigate away somehow... and I loved this video. Well done. You've won me over,.
NFTs and Crypto Currency never impressed me, but AI has already shown it has great potential. Sure, not all companies will be successful and many will fail, but AI is here to stay
What can it do ?
@@piccalillipit9211 he said it has potential, like, look how generative AI couldnt do hands and now it can, or how it can simulate a human voice that hears so so real with laughs and breathe pauses, etc (chat gpt-o), how we could never imagine to generate a video from text and then it does, etc... none of it its perfect but it has a lot of potential if it reaches a nice level of quality
@@gatolocoowo Yeah - great. What can it do? what use case does it have? Cos hallucinations are a FEATURE that will never ever go away. So that means it cant be used in any regulated industry. It cant fly a plane cos X% of the time "crash into the mountain" will be a legitimate action, it cant be used in finance cos "lose your client money" is a legit probability - it has not idea of good / bad or right and wrong.
It is NOT Generative Artificial intelligence at all - Its a Generative Probability Engine.
so it can only be used in applications where "machine does bad thing" is acceptable. Cos there are an infinite number of bad things, and it WILL do them
ALL use cases are predicated on it NOT being what it is.
LLMs and other ML models? Sure. They're here to stay and I can say that confidently because they're frankly nothing new. In the hands of trained professionals they can be game changers, especially in science. But not everything needs an "AI" chatbot to give you confidently wrong answers - especially when it costs an arm and a leg (and a fuckton of carbon emissions) to run.
technologies can be incorporated just as much as they can fall out of favor. If in the end the costs will be to great to possible profit - AGI is a gambit because we know it to be a sci-fi dream - we'll see for how long it is here to stay
This deserves way more views.