5 ms·

I'm not sure to what degree large scale LLMs are a bubble. My sense is there are enough useful applications lurking in there that people will be paying ChatGPT

by TimPC 3y ago

I'm not sure to what degree large scale LLMs are a bubble. My sense is there are enough useful applications lurking in there that people will be paying ChatGPT API fees for a long-time. I also think those fees are high enough that the company is not only covering costs but paying out huge bonuses to every developer. As for other AI technologies, we already have many proven examples in use today generating value. Things like automated text scraping are everywhere. Applications like TTS and ASR are somewhat common-place as well. I think the big difference between AI and previous tech bubbles is that AI has a huge amount of useful research literature demonstrating various kinds of value underlying it. There is a question of whether research value can be translated into business value effectively, but there are very real improvements in AI that underlie the explosion in AI businesses. I'm hard pressed to find the same level of technical innovation in crypto over the same time period. As I understand it most cryptocurrencies run minor variations of one of two algorithms. I also think the dotcom bubble in general had very little of the same research underpinnings. There are a lot of AI problems where we are twice as good as we used to be five years ago. It's not surprising that some start-ups will exist to check if that's good enough to generate business value. I'd be very shocked if this AI Bubble didn't leave a lot of good things behind because I think it's very unlike the other bubbles in the fundamentals that underpin it.

ska 3y ago
> My sense is there are enough useful applications FWIW tech bubbles always have a large number of useful applications, hell the dot-com bust was full of companies held up for mockery that (maybe?) turned out just to be 15 years to early. Bubble doesn't mean "there is nothing useful there" it just means the interest/focus/investment gets unmoored from the fundamentals. I think we are pretty clearly there in the AI space right now, even if you focus on core technologies that are working. >very shocked if this AI Bubble didn't leave a lot of good things behind you are basically arguing it's his "Bubble 1.0" rather than "Bubble 2.0", and he is less optimistic, but you aren't arguing fundamentally different things.
rented_mule 3y ago
I don't think a bubble associated with a technology means that the technology is not incredibly useful. Rather, it's an overly aggressive inflation of value that is nominally attributable to that technology, where much of that value later collapses. The dot com bubble did not invalidate the importance of the web. But there were all kinds of shenanigans going on around the promise of the web. People were setting up companies thinking/claiming that making a website would make the business wildly valuable, ignoring all the actual problems unrelated to the web that they'd need to solve before they would have viability. They were then successfully going public with nothing but the website built - no actual product in place. Then it collapsed. But the web carried on and some of the companies being built through that period (e.g., Amazon, eBay, Google) were quite real. I see similar things going on with AI now... An electric utility company that a friend works at has an edict to lay off all 200 customer support staff and replace them with a chatbot. They don't want to pay OpenAI, so the executives are giving the team (of 3 data scientists) 6-months to build their own in-house LLMs. They don't have the data or the infrastructure expertise to do it. But, AI! A tech company with 1-2K employees and lots of revenue that another friend works at is preparing to IPO in the next year. Executives are forcing LLMs on teams that can't find reasonable applications for them. It's being done to boost their valuation heading into the IPO because they can then call themselves "AI powered". At a FAANG that another friend works at, a non-technical VP with lots of available budget commandeered several data science and engineering teams to build something laughably bad with LLMs (LLMs would never work for this application, customers would hate the outcome, and it would create contractual problems for the company). The technical teams were tempted by the budget after a round of layoffs, but they could only stomach it for so long. After most of a year and 10s of millions of dollars spent (plus the contractual issues starting to appear), sanity prevailed, and the project was cancelled. None of these examples say that LLMs (or other ML approaches) won't continue to have life/society altering impact. They point to people trying to exploit/build the hype for short term financial gain. It's why so many ML practitioners don't like the term AI - many of them feel like AI == ML + hype. ML is just a tool, and all the hype makes people engage in science-fiction type thinking that obfuscate where the value really is. But it will all work because AI! These are only anecdotes, but what I'm seeing now feels a lot like what I saw in the dot com bubble.

by TimPC 3y ago

ska 3y ago

> My sense is there are enough useful applications FWIW tech bubbles always have a large number of useful applications, hell the dot-com bust was full of companies held up for mockery that (maybe?) turned out just to be 15 years to early. Bubble doesn't mean "there is nothing useful there" it just means the interest/focus/investment gets unmoored from the fundamentals. I think we are pretty clearly there in the AI space right now, even if you focus on core technologies that are working. >very shocked if this AI Bubble didn't leave a lot of good things behind you are basically arguing it's his "Bubble 1.0" rather than "Bubble 2.0", and he is less optimistic, but you aren't arguing fundamentally different things.

rented_mule 3y ago

I don't think a bubble associated with a technology means that the technology is not incredibly useful. Rather, it's an overly aggressive inflation of value that is nominally attributable to that technology, where much of that value later collapses. The dot com bubble did not invalidate the importance of the web. But there were all kinds of shenanigans going on around the promise of the web. People were setting up companies thinking/claiming that making a website would make the business wildly valuable, ignoring all the actual problems unrelated to the web that they'd need to solve before they would have viability. They were then successfully going public with nothing but the website built - no actual product in place. Then it collapsed. But the web carried on and some of the companies being built through that period (e.g., Amazon, eBay, Google) were quite real. I see similar things going on with AI now... An electric utility company that a friend works at has an edict to lay off all 200 customer support staff and replace them with a chatbot. They don't want to pay OpenAI, so the executives are giving the team (of 3 data scientists) 6-months to build their own in-house LLMs. They don't have the data or the infrastructure expertise to do it. But, AI! A tech company with 1-2K employees and lots of revenue that another friend works at is preparing to IPO in the next year. Executives are forcing LLMs on teams that can't find reasonable applications for them. It's being done to boost their valuation heading into the IPO because they can then call themselves "AI powered". At a FAANG that another friend works at, a non-technical VP with lots of available budget commandeered several data science and engineering teams to build something laughably bad with LLMs (LLMs would never work for this application, customers would hate the outcome, and it would create contractual problems for the company). The technical teams were tempted by the budget after a round of layoffs, but they could only stomach it for so long. After most of a year and 10s of millions of dollars spent (plus the contractual issues starting to appear), sanity prevailed, and the project was cancelled. None of these examples say that LLMs (or other ML approaches) won't continue to have life/society altering impact. They point to people trying to exploit/build the hype for short term financial gain. It's why so many ML practitioners don't like the term AI - many of them feel like AI == ML + hype. ML is just a tool, and all the hype makes people engage in science-fiction type thinking that obfuscate where the value really is. But it will all work because AI! These are only anecdotes, but what I'm seeing now feels a lot like what I saw in the dot com bubble.