# Your all Wrong...AI Is NOT A Bubble!

## Метаданные

- **Канал:** TheAIGRID
- **YouTube:** https://www.youtube.com/watch?v=T3B4AesBRuo
- **Дата:** 24.07.2024
- **Длительность:** 22:10
- **Просмотры:** 24,717

## Описание

Prepare for AGI with me - https://www.skool.com/postagiprepardness 
🐤 Follow Me on Twitter https://twitter.com/TheAiGrid
🌐 Checkout My website - https://theaigrid.com/


Links From Todays Video:


Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.

Was there anything i missed?

(For Business Enquiries)  contact@theaigrid.com

#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience

## Содержание

### [0:00](https://www.youtube.com/watch?v=T3B4AesBRuo) Segment 1 (00:00 - 05:00)

so one of the highly debated topics right now is whether or not AI is currently in a bubble since there has been the explosive growth in generative Ai and certain stocks have rocketed to stardom many people are speculating that AI is a bubble combined with the fact that many of the pitfalls of generative AI are unlike any technology that we've ever seen some people are beginning to question ai's economic value and whether or not this entire thing is a house of cards currently on screen you can see two popular videos released by YouTubers in the finance SL information space theyve released I wouldn't call these hit pieces but videos in which they document why the AI bubble is bursting now in this video I'm going to of course firstly talk about some of their points and not just these videos but recent reports from Goldman Sachs and other investment firms speaking about the pitfalls of generative AI I do want to say that this is something that I actually disagree with despite potentially having a biased opinion as someone who does an AI Channel I will say that there are many different factors that many people are simply overlooking which just absolutely doesn't make sense I'm going to dive into the details and let's get into why AI is most certainly not in a bubble now one of the things that they keep on referencing is the dotom bubble the dotcom bubble also referred to as the internet bubble refers to the period between 1995 and 2000 when investors pumped money into the internet-based startups in the hopes that these fed L companies would soon turn a profit during this time companies skyrocketed their valuation just based on the fact that they added Doom now of course one thing that was with the dotom bubble that many people are continually forgetting is that the internet actually did bring about tons and tons of economic value to the world and so I think comparing these things without remembering the fact that the Doom bubble despite its ability to have some comparisons did actually yield a true economic gain across many different sectors in Industry there was also a recent report from Sequoia they've called it ai's $600 billion question the AI bubble is reaching a Tipping Point navigating what comes next will be essential now this is one of the main questions that they do talk about I'm going to show you guys they talk about where is all the revenue essentially I'm going to read this small part here it says at that time in September 2023 I noticed a big gap between the revenue expectations implied by the AI infrastructure buildout and the actual Revenue growth of the AI ecosystem which is a proxy for end user value I describ this as a $125 billion hole that needs to be filled for each year of capex at today's levels the question on where is all the revenue in terms of the big gap between revenue expectations and of course the AI buildout it's a question of understanding the fact that what many people are investing in is not current AI expectations but if we take a look at the next 5 to 10 years of AI when true transformative AI is expected to be here those capable AI systems are very likely going to be capturing large amounts of economic value one thing that I've seen time and time again from many people publishing reports is that they can't seem to Fathom why companies are spending billions of dollars on AI buildout infrastructure if there isn't any current economic value but what many people haven't realized and I'm not sure why this is that many companies are betting on the late stage AI or as I would put it companies are not actually betting on AI they're betting on AGI which many people are overlooking and I'll dive into this further later now there was also another article which covered Jim cllo from Goldman Sachs who says that he believes that AI will cost trillions and doesn't believe that it will fundamentally change lives he says that AI technology is expensive and inaccurate and that's very unlikely to change you can see here it says but for now CLL remains a skeptic about the truly transformational AI use cases I haven't seen any yet and I don't anticipate there's going to be any yet I mean I'm not trying to be someone who is just a complete AI bull or a complete AI Doomer but to say that there are no transformational AI use cases is just completely false a quick look at the GPT 40 demo and we could clearly see that even a voice AI system is going to provide many different use cases that are going to be used across a wide range of Industries now I guess you could argue that these might not be truly transformational but you have to remember adoption type right now we are still in the early stages of adoption to where there is still that question of should we actually use these technology you have to understand people don't just start using technology all at once there are the people who are early there of course are the innovators the early adopters the Early majority the late majority and of course the lagard now this statement right here I have to say that I disagree with not seeing any

### [5:00](https://www.youtube.com/watch?v=T3B4AesBRuo&t=300s) Segment 2 (05:00 - 10:00)

transformational AI use cases and stating that there's not going to be any I mean it just seems like that person isn't paying attention and I know that might be crude for me to say that but I just truly don't understand where that statement even comes from it continues to state that the bank's head of equity Global Research coell is a technology industry veteran who has witnessed 30 years of transformative shifts such as the invention of the internet the e-commerce Revolution and the adoption of smartphones and much more I put this in here to show you guys that this person isn't by any means someone who doesn't have the experience to understand transformative technology but I think what this does show us is that if someone with this much experience is underestimating the capabilities of AI it's a trend that we've seen time and time again and that trend is one of underestimating ai's capabilities is something that you never want to do because all it takes is one breakthrough one research paper or one transformative Discovery to revolutionize what the landscape understands and of course the coming application now one of the things that has been said about this is that this is too expensive as previously sooya said the $125 billion question or now the $600 billion question is one that many people are thinking all of this expenditure and for what you can see here that the potential use cases of these Technologies were Apparent from their Inception and this is in reference to the internet and of course the telephone but cavello sees no comparable road map for applications of AI today at best C's AI is a tool that can make existing tasks easier or more efficient and at worse it's hallucinating virtual assistant and of course this is referring to Google's Pizza glue where if you asked how to make pizza stickier with cheese they recommended glue because it essentially pulled data from Reddit now this doesn't really make sense saying that AI is only going to make tasks easier or more efficient and at worst be a hallucinating assistant doesn't factor in all of the things that AI is going to do considering the rate of improvement in the future but anyways let's tackle the first thing the point here that I was actually meant to reference was the fact that cavello points out the AI use cases replacing lowcost jobs with very costly technology the opposite of other developments he's seen and here he says most technology transitions in history have been a less expensive solution replacing a costly solution he points out the e-commerce Revolution where Amazon was able to rapidly take market share away from traditional brick and mortar retailers due to the lower storefront cost however I don't think he's also factored into this you see I also read Goldman Sach report and this was rather interesting you can see here that someone also comments on the too expensive fallacy being just that a fallacy he says that AI technology is undoubtedly expensive today and the human brain is 10,000 times more effective per unit of power in performing cognitive tasks versus generative AI but the Technology's cost equation will change just as it always has in the past in 1977 in 1997 a sun microsystem server cost $64,000 within 3 years the server's capabilities could be replicated with a combination of x86 chips the point here is that whilst yes generative AI is currently expensive we know that once technology is being built usually it is always going to be quite expensive and a lot bigger than it needs to be if we look back at previous computers those things were huge I mean if you look at the computer SL thecode that put the people on the moon those things were very inefficient and if we look at mobile phones and TVs and how much those have slimmed down and changed in the past we can see that technology usually gets rapidly more efficient and rapidly more cheap and even recently there was GPT 40 mini you can see here that Sam mman talks about how way back in 2022 the best model in the world was tex3 it was much more worse than this new model and it cost 100 times more so essentially if we look at gpt3 costing 100 times more than the current model now we can see that intelligence is currently decreasing whilst yes the build outs are still very expensive we are still headed in One Direction now another thing that cavello points out and I just have to disagree with this as well because I've been paying attention to the research papers and all of the development surrounding Frontier models he says that to those who argue that AI will become smarter through further training cavello points out that these models have already consumed copious amounts of data and are quite far along in their training in fact big Tech is rapidly running out of data to train it on now this is a good point and I'm going to continue this point let's just take a look at this he says technology that is highly skilled at analyzing historical data but unable to apply it to new situations won't dramatically change our everyday lives or replace any high value jobs if AI doesn't actually have the ability to conduct higher order reasoning it won't leave a trillion dollar impact on the economy now let's actually break down this statement okay one of the first things that he said here was that these models have already consumed copious amounts of data and are quite far along in their training that is a TR statement however Frontier labs

### [10:00](https://www.youtube.com/watch?v=T3B4AesBRuo&t=600s) Segment 3 (10:00 - 15:00)

are already differing in how they actually utilize these models with new techniques for example a growing approach is neuros symbolic AI this is where we use large language models with different kind of agentic architectures or with tools or with different sapling methods to get AI systems that are much more robust in terms of their reliability and their reasoning capabilities this means that when we have future models experimenting in different training methods we're very likely to get systems that are much smarter than ourselves one of the key callbacks that we can continue to mention is Alpha go it was a system that was able to get truly extraordinary abilities just based on human level data but once it trained on synthetic data where it played against itself it was truly able to surpass all humans in its environment and I do believe that once these Frontier labs are starting to focus on that area that's when we're going to see the increase in reliability and of course the increase in how these systems perform but what this also doesn't take into account is the fact that there are still two more training Cycles before we actually exhaust these kinds of capability so stating that they're quite far along in their training doesn't account for the next two orders of magnitude to which we will get a lot more reliability and a lot more robustness in terms of what these models are already allowed to do and what this also doesn't take into account is the agentic architectures that are built around these models that give us an additional 10 to 25% in terms of performance increases now this second slide here is pretty self-explanatory if AI doesn't have the ability to conduct high order reasoning it won't leave a trillion dollar impact on the economy now if you've even been paying any kind of attention to the space one of the things that everyone is now focused on is higher order reasoning people that Google Deep Mind and the folks are open Ai and at anthropic their main focus is of course conducting reasoning because this allows the model to be applied to many different situations including novel scenario now recently open ey even spoke about how they were approaching the stage of reasoners and how their next Frontier model GPT 5 is very likely to be at that of a PhD which means that as these models increase in their capabilities as the scale increases and as we figure out Rays to conduct high order reasoning which we know is going to happen a trillion dollar impact on the economy is most certainly going to be likely now here is Bill Gates also talking about the AI bubble and comparing it to the 1990 Fiasco when we had the internet bubble multiples aren't as high as they were say during the internet bubble uh um and the growth is real I mean AI is not pets. com this is something very fundamental so I you know I I'm not don't take my view on any particular stock uh but the growth potential you know there will be some big winners in this AI space which is why you see all the leading tech companies including Microsoft putting tens of billions of dollars into not just the backend capacity but re-engineering their application so you're your far more productive so you think that the excitement that we're seeing in these valuations is fully warranted by the potential the excitement we're seeing absolutely any particular stock I'm not really commenting on but we're not out of control we're seen a uh fundamental Advance as important as any in the history of digital technology Bill Gates like many have a different approach because I think he understands that whilst yes there is a lot of hype it's not hype for any reason the technology is truly transformative and I think the majority of people who are in this space are not looking at what it can currently provide but looking forward to seeing where if there are only a few more breakthroughs the technology is going to be truly transformative many of those who are analyzing and looking at current data and current expenditure and current applications are failing to realize the long-term effects of what this technology is able to do now in Goldman sachs's actual report there was also another thing here we can see that it talks about how skeptic miss these three things Eric Sheridan says again I readily acknowledge that the return on invested Capital visibility is currently low and the transformative potential of AI will remain hotly debated until that becomes clearer but AI Skeptics Miss three key things one training on existing historical data to inform and drive analytical outcomes in the future exactly sounds like going to University people go to learn and then improve productivity and efficiency for decades after graduation and machines can absolutely do the same number two machines today can do a whole host of tasks more productively and more efficiently than humans and that will remain true for decades into the future and three people didn't think they needed smartphones Uber or Airbnb before they existed but today it seems Unthinkable that people ever resisted such technological progress and that will almost certainly prove true for generative AI technology as well now one of the things I always wanted to emphasize in video is that AGI is

### [15:00](https://www.youtube.com/watch?v=T3B4AesBRuo&t=900s) Segment 4 (15:00 - 20:00)

completely different to ai of course is something that has a variety of different use cases and many different applications from narrow AI to self-driving AI to algorithmic AI but I think AGI is going to be the main focus of these Frontier labs and I think that is completely different to the standard AI that they might be thinking about the reason I'm talking about AGI is because the economic value of AGI is in the tens of trillions of and some are speculating it could capture up to 10% of the world's GDP this is because if you truly understand what an artificially general intelligence system is able to do is able to pretty much do any task better than any human and that would seriously unlock a huge level of economic value for those that own it now here's Sam Alman in an interview basically talking about the fact that these expenditures don't really matter because they're going to build AGI whether we burn 500 million a year or 5 billion or 50 billion a year I don't care I genuinely don't as long as we can I think stay on a trajectory where eventually we create way more value for society than that and as long as we can figure out a way to pay the bills like we're making AGI it's going to be expensive it's totally worth it and the reason Sam mman has said that it's totally worth it is because if you can actually get to AGI you are truly going to be reaping the rewards of that like no other company has ever seen and you have to understand as well is that these giant companies there's only three or four of them that can really compete at the frontier skill you've got anthropic meta Google open AI Apple if they want to play in the game but there's not too many different players here and we haven't really seen insane valuations that aren't backed by actual revenues of these companies even Nvidia actually has the billions of dollars due to the fact that there has been increasing investment into the space now one of the things that people need to understand is that this stuff isn't decades away it's only 3 years away if you've been paying attention to situational awareness the intelligence explosion section in which we can see how future AI systems are going to get increasingly more capable we know that around 2028 to 2027 there is likely to be the intelligence explosion or that will be the area in which many predict we do get AGI now when I'm talking about predictions for AGI one thing that I've realized and come to notice is that these predictions are not made by people who are hyping up products so for example in some Industries some people predict that you know this technology is going to be here so that they can drive more investment into the space a lot of the predictions that I'm seeing about AGI are from actual esteemed researchers who are at these Frontier companies leading the way in terms of what these systems are capable to do which means that their statements aren't to be taken lightly their statements are backed by Decades of research and looking forward on what's probably to come now this chart right here is by someone who worked on super alignment at open Ai and he said that this isn't something that is out of the question believing in super intelligence isn't something that you need faith in it just needs you to look at lines on a graph and you'll see that super intelligence is most certainly possible one of the key things also that people need to remember is that the only thing that these Frontier Labs need to do is automate AI research and then everything from there on out is quite easy the reason I say that is because once you automate AI research then the company's automated itself and everything off the back end is just going to be pure profit in terms of what you're going to be able to do with an increasingly powerful system now here we have James beta basically saying that we're only 2 to 3 years away from system to thinking which is this incredible level of reasoning that's going to be able to do other things I did cover this in the video but I'm going to read you guys a too long to read in summary we've basically solved building World models have two to three years on system two thinking and one to two years on embodiment the latter two can be done concurrently once all of the ingredients have been built we need to integrate them together and build the cycling algorithm described above I'd give that another 1 to two years so my current estimate is 3 to 5 years for AGI I'm leaning towards three for something looks at an awfully General intelligent embodied agent which I would personally call an AGI then a few more years to refine it to the point that we can convince the Skeptics of the world so that is an open AI research engineer and not only did he State this but there is also someone who left deep mind because he believes that AI is only 3 years away there was also Mustafa sullan the co-founder of Deep Mind and currently the CEO of Microsoft's AI team he says certainly before the end of the decade we are not just going to have those capabilities but those capabilities are going to be widely available for very cheap potentially even in open source and he said that potentially this Tech will be able to run a business within the next 5 years and AI might be able to act like an entrepreneur and an inventor we can also see that Sam Alman has said that he's question on when he thinks AGI will be a reality it's 5 years give or take maybe slightly longer but no one

### [20:00](https://www.youtube.com/watch?v=T3B4AesBRuo&t=1200s) Segment 5 (20:00 - 22:00)

knows exactly when or what it will mean for society we also have Dario amod the CEO of anthropic talking about when he believes AGI will be achieved like a human Child Learning and developing they're getting better and better smarter and smarter more and more knowledgeable in 2025 2026 maybe 2027 there is in my mind a good chance that by that time we'll be able to get models that are better than most humans at most things what is catastrophic risk so considering the fact that we have the majority of AI researchers and the majority of CEOs at Frontier Labs basically saying that we've got 3 to 5 years before AGI most of them leaning towards 3 years and considering the fact that if we look at the data and we pay attention to the graphs of where super intelligence and these truly capable systems are going to be while yes the spending initially is going to be expensive I think longterm this a clear technological Revolution that most people are overlooking if AI were to be in a bubble whilst yes if you do look at the private Equity area where money is being raised in the millions and millions of dollars it isn't being raised at a stage to where you can say that the valuations are absolutely crazy and in order for it to be a bubble there would need to be tons and tons of valuations with little to no Revenue so with that being said I would have to say that I completely disagree that AI is in a bubble and whilst yes these videos do make for enticing clickbait headlines and staing that AI is dying I just think there is the trough of disillusionment in the Gartner hype cycle now I think the AI hype is going to be there once again when the next round of Frontier capability systems are released such as Gemini 2 claw 3. 5 Opus and of course GPT 5 but as someone that's in the space I can say that when AI does become a bit bubbly you'll be the first to know about it on this channel whilst yes you might think this is completely biased because my channel is about AI trust me when I tell you I can remain completely unb biased because it wouldn't make sense for me to continue to invest in something if it wasn't a bubble so with that being said if you enjoyed the video I'd love to see your thoughts and comments in the comment section below and hopefully have a great rest of your day

---
*Источник: https://ekstraktznaniy.ru/video/14171*