The Coming AI Startup Bust #finsubito agevolazioni

1. VC math revolves around power laws. It doesn’t seem far fetched that long term we end up with a couple trillion dollar AI companies – we’ve already seen this play out in the past. If that becomes true than these valuations are not THAT far fetched. Obviously that’s a big if though.

2. @10:20 dutch tulips from 1636

3. AI is so over hyped right now I'm kind of hoping the bubble would burst sooner rather than later. To be clear there absolutely are practical use cases for AI and I expect more will be discovered as startups come and go. Still the amount of absolute shovelware products with "AI" sticker slapped on them is clear sign that there indeed is a bubble.

4. I have a hard time swallowing optimistic words when they're about AI. I'm generally not in favor of LLM technology or the modern image recognition and generation technology. And as an advertising pitch, it makes me choose not to buy something. And I think this has been the case for other people.
   I will never buy an AI product unless I physically have no other alternative.

5. I worked in an AI lab in the 1990s, back then AI was seen to be in a crisis as what AI could do had hit a wall which nobody could mount. AI then sat for 25 years and then suddenly everyone was putting money in people who were using 30 year old methods on massive clusters of processors. Maybe all the AI PhDs I knew in the 1990s were wrong and Silicon Valley investors will be correct. Maybe not.

6. LLM are facing a limitation that I would argue is inherent in their models. Right now the models are not improving as more CPU and time is thrown at them, the LLM has hit a bit of a wall. This is a start up industry going into unproven technology. So we are in the same place as self driving cars, vertical farming, Social Media and lots of other techs that are getting masses of money while not sure that the tech will ever work.

7. a

8. Do not forgot about the bitcoin bubble 😂

9. Develop AI for lights-out manufacturing, precision farming and open source intelligence

10. I do see a hard crash on AI company stocks in the US in the coming years; it's a heavily speculated business.
    However I see Chinese AI companies doing okay with sustained state investments and will probably outgrow the American AI companies after a decade or so.

11. Silicone Valley had the same exact problems. Issue isn't the technology, the issue is marketers and social ramifications.

    Google took a look at programmers in their garages and went "hey, you know what increases productivity? FLUFF!'" and proceeded to create a culture strictly related to style over substance (the ping-pong tables, beanbag chairs, bright decorum, etc.).

    Turns out, suits don't know shit about programming and, as it turns out, it doesn't stop there: "forget innovating new stuff, let's just re-arrange the GUI on YouTube for the 50th time to keep people 'engaged' 😀 "

    So, in summary, no, as it turns out, you cannot take a bunch of retarded suits that know nothing about programming or technology and, no, you *cannot*, in fact, innovate nothing by turning your programmers into an art project you can show off to OTHER suits and go "lookie what I own".

    Google, Microsoft, AI?

    They're all run by suits, not mathematicians; they're all concerned with how to privatize $0 worth of effort (and trademarking it so nobody else ever has a chance to make the same thing), not creating better tech.

    This won't stop until technology stops being a business and starts being about individual problem solving.

12. At some point in the near future, VC will dry up and profit expectations will rise dramatically. That is when we will be asked to pay for all of the currently free AI BS. Mostly grift, partially functional, and hardly revolutionary. Let’s see if I am correct. Two years.

13. I am working on an AI startup myself and I think you are right. It's very easy to copy each other. I believe that the ultimately determining factor will not be the product itself, but the business operation behind it

14. 9:52 – "Brings prosperity across the classes" what are you high on? Share me some of that, because that is utterly crazy. The most likely scenario is that thousands might lose their job and a huge portion of the middle class simply disappears, with all of them going to lower income and shitty jobs. And at the top only a very few will be successful, just like the hard drive boom, or the real estate boom. Only the larger survive, absorbing the smaller ones in their wake. We have no reason whatsoever to believe that AI will bring prosperity.

15. it'll be ok, you say? so all the AI will come crashing down and never come back?? whew, what a relief

16. VCs are greedy C students driven by FOMO. The lucky ones get one hit to make their money and then they just collect the next round of greedy C student money. I’ve been in tech for almost thirty years and lived through IPOs, booms and busts already, particularly cyber security. It’s wild but predictable. Great video as always.

17. nope. You get the profits side completely wrong.
    AI boom is very similar to search engine boom. Search engine boom was similarly perplexing and many predicted that companies like google – offering search engine free of charge – will quickly collapse – because where is the profit?
    Google data centers seemed giant and unsustainable. Yet there are two major revenue sources everyone missed :
    1)userbase. Successful searches attracted more users. Those users remained as resource – you can offer them other services as they stay on your platform as long as your platform remains useful. Then you hook them even more – offering mail, maps, calendar etc.
    and this leads to another giant revenue driver
    2) what people actually type in.
    That's a real gold mine. This together with metadata provides you endless free data. Just to remind you – the majority of AI training data which made AI possible at all comes from what people typed into search boxes, blogs , email, gboard keyboard, maps and map movement data , etc.
    Information on what people search the most, the least, information about unusual searches (f.e. users trying to find products not yet existing), even typos – are all valuable assets to re-sale.

    AI is no different but on steroids – it actually understands users, building so called "human terrain system" able to create real-time live maps of society, its problems, desires, plans.
    This provides invaluable marketing information – f.e. allowing big corporations to not only know single keywords people used for search, but also what sophiscation users language is, what is their cultural literacy, what are their coding skills, higher and lower language comprehension level, linting and debugging skills, and last but not least – hardware and system capabilities.
    No longer major corporations like Microsoft have to speak to drunken geeks on the Usenet searching for answers to the question "what language Microsoft should use to write its new compiler for" .

    Last but not least remember the
    3) All stuff AI produces is owned by AI vendor..
    That means every line of code, every poem, bucket list – all cannot be really re-licensed. If someone creates great algorithm using chatGPT – chatGPT owners own it and by owning i mean not only license rights, but the very fact they have it on their servers and can do whatever they wish with it (f.e. re-sell it or integrate into their own product) That's huge.

18. AI is notoriously whacked. Depend at your own peril.

19. Oh please let the bubble burst soon. I work in software development and am so fed up about everything has to have AI, which just means some LLM shoehorned in somewhere ahh 🤢

20. You are a smart man. You are not incorrect. Even i can foresee this.

21. Could you cover photonics AI companies like "Celestial AI" and "Lightmatter”?

22. Here's more proof that it's a bubble: Look at all the comments on this video defending the AI bubble which neither offer anything substantive to back their faith in it nor demonstrate any understanding of what LLMs are or how they work.

23. anyone else work for data annotation? … watching to figure out when will everything dry up lol

24. Wait till you hear about Tesla.

25. Bubbles bubbles always something

26. All AI should be banned.

27. Ironically you state this on stg called the internet which was a bubble from 1998-2000 and bursted massively in 2000

28. For anyone joining a startup, you know there is a huge risk in joining. That is on THEM to manage their finances well. If you don't want exciting work for a larger company.

29. Cyrpto tech hype, now we have ai hype.. pathetic .. but more pathetic is laying off people for an ai, ceo just want to save money by overworking remaining staff

30. You're looking at this from an ivory tower. Ordinary people will lose money and jobs, and very little of value will be created.

31. All that crap is NOT OK. It is all funded by speculation and without fiat currencies it would not be possible at those levels. .

32. Bruce Wayne on the AI bubble: let's get nuts!

33. Some bubbles just don't burst. Cope harder.

34. I can almost see myself as the boomer soon enough that insists that AI tech is just bs for the big oligopolies that will manage to survive the eventual burst lol. Boom and busts are how the modern day economy runs, personally I think AI will never 'burst' but will eventually just have a slowly receding growth similar to Japan's lost decade – never really exploding but fettering out all the same.

    AI brings in real value and even if its in such a highly competitive space, the fact that governments, specifically military-industrial complexes are trying to develop it themselves should keep it afloat long enough for unprofitable startups to be absorbed.

35. Why are you speaking so slowly?

36. This is the time to get messy. Brilliant!

37. I'm not convinced the current historical moment is ideal for "capitalist exuberance", honestly. Could we try capitalist humility for once?

38. AI is ruining the internet. Now almost everything I stumble upon is fake or terribly made.

39. It’s a bubble….everybody knows it. Hopefully some very rich people take a bath….

40. 20% of index funds are compromised of the tech sector and 40% of market returns in the last 15 years are from the tech sector. Grandma will lose her money.

41. It's all so stupid and tiresome.
    My s23 ultra is supposed to have some AI shit, I've never used it in my life, usually it only gets in the way of stuff I've try to do since it's attached to weird gestures.

42. AI startups are very risky. The current SOTA models all require extensive training resources to achieve their performance. At least based on recent papers, recent performance gains are mainly due to increasing model size.

    But to achieve this performance, a huge amount of data is required for pre-training, as well as human feedback for fine-tuning. No start-ups will have the kind of resources required to pull this off from scratch. It will cost a lot of money to buy/rent GPUs, hire 3rd parties for data curation, RLHF, etc. 100 million is peanuts.

    In the end the winners will be the mega corps like Meta, Google, Microsoft who has the capital to run huge GPU clusters and enough money to push models to new SOTA.

43. My AI powered toilet uses adanced machine learning to flush your poop as soon as it drops into the water.

44. In order to build something you need talent. No one is creating talent. They want to pull experience from other companies that have already developed experienced talent. So there WILL be a shortage in talent. There isnt nearly enough power to self-sustain this without humans. Furthermore, we will see government restrictions eventually as these companies have free range to build potentially dangerous platforms.

45. As a time traveler, this video aged like milk.

I commenti sono chiusi.