I don't think companies like Nvidia, OpenAI, Microsoft, Apple, Google, Meta are in a bubble. They're making real profits or they are foundational. Likewise, I don't think S&P500 companies who are adopting LLMs in a big way are in a bubble.
I do think startups with multi-billion dollar valuations but mostly relying on OpenAI/Claude are in a bubble. There's no reason a company like Windsurf was worth $3b or Cursor should have $10b.
bayindirh•5mo ago
While I make a lot of anti-LLM comments here, my qualm is with user facing "AI companies" using internet as their scrape-ground to detriment of everyone.
From what I see, the real AI, the ones users don't see, but the ones powers research infrastructures or working on private data are doing some real work and creating real results is the future. The LLMs many see are toys when compared to these ones.
I have seen AI models run experiments in lab, the boring vision models like YOLO creating real predictions and insights about our ecosystem, and fine tuned local LLMs provide real insights.
Hardware companies are not in a bubble because of this, but user facing LLMs are in a bubble.
goalieca•5mo ago
What happens when LLM companies stop funding their services at a loss? The downstream parts and services will stop seeing revenue. Those people are making a killing off AI since they are so expensive to build and run.
aurareturn•5mo ago
They're already profitable per token in terms of inference.
They're losing money because of R&D and spending capex to buy GPUs.
AbstractH24•5mo ago
The fact this comment isn’t gaining more traction and raising more flags is itself one of the biggest indicators we’re in a bubble to me.
hellisothers•5mo ago
I realize this is suspicious tin-hat in a can but it makes me wonder what he and OpenAI have to gain by him saying this? Does he feel they got their nut and now it’s time to undermine the ability of other AI companies to raise money?
cholantesh•5mo ago
In the article he specifically shitcans AI startups that were founded by former OpenAI execs, but also he has a habit of characterizing AI as something that needs expert hands, mainly his, to manage and contain. At the end of the article he delivers on this self-aggrandizement.
goalieca•5mo ago
The leadership of the big AI companies have made some pretty extreme promises on timelines and capabilities. I wonder he’s changing his tune because he’s worried investors will sue him once the man behind the curtain is revealed.
reactordev•5mo ago
doubtful. I think this is a classic case of Sam downplaying it so that others get discouraged and exit the market. He's stated a few times that the improvements from GPT-3 to GPT-4 were "minimal" when they clearly weren't.
I also think they intentionally hamstrung GPT-5 to back this claim so that they can slow down and focus. Just my gut feeling though and I have nothing to back my claim.
IAmGraydon•5mo ago
Who knows, but one thing that always holds true about Sam Altman is that what he said is not to be trusted. From all the way back to his days at Reddit, he's shown himself to say whatever is necessary to win. His recent comment struck me as strange because if there's one person who is always going to great lengths to pump the bubble up as much as possible, it's Sam Altman.
aurareturn•5mo ago
I do think startups with multi-billion dollar valuations but mostly relying on OpenAI/Claude are in a bubble. There's no reason a company like Windsurf was worth $3b or Cursor should have $10b.
bayindirh•5mo ago
From what I see, the real AI, the ones users don't see, but the ones powers research infrastructures or working on private data are doing some real work and creating real results is the future. The LLMs many see are toys when compared to these ones.
I have seen AI models run experiments in lab, the boring vision models like YOLO creating real predictions and insights about our ecosystem, and fine tuned local LLMs provide real insights.
Hardware companies are not in a bubble because of this, but user facing LLMs are in a bubble.
goalieca•5mo ago
aurareturn•5mo ago
They're losing money because of R&D and spending capex to buy GPUs.