Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

foundationcapital.com

Post reply on HN
Print view View original

101–110 of 138 posts

Re: Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

#101

Post by mirzap »

mbowcut2 wrote: post #86

DeepSeek has demonstrated that there is no technical moat. Model training costs are plummeting, and the margins for APIs will just get slimmer. Plus model capabilities are plateauing. Once model improvement slows down enough, seems to me like the battle is to be fought in the application layer. Whoever can make the killer app will capture the market.

Model capabilities are not plateauing; in fact, they are improving exponentially. I believe people struggle to grasp how AI works and how it differs from other technologies we invented. Our brains tend to think linearly; that's why we see AI as an "app." With AI (ASI), everything accelerates. There will be no concept of an "app" in ASI world.

Re: Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

#102

Post by abathur »

mbowcut2 wrote: post #86

DeepSeek has demonstrated that there is no technical moat. Model training costs are plummeting, and the margins for APIs will just get slimmer. Plus model capabilities are plateauing. Once model improvement slows down enough, seems to me like the battle is to be fought in the application layer. Whoever can make the killer app will capture the market.

Can you unpack why you think there'll be defensible moats at the application layer?

(I thought you had this exactly right when I read it, but I kept noodling it while I brushed my teeth and now I'm not so sure llms won't just prove hard to build durable margins at meaningful volume on?)

Re: Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

#103

Post by tarsinge »

keithwhor wrote: post #93
mbowcut2 wrote: post #86

DeepSeek has demonstrated that there is no technical moat. Model training costs are plummeting, and the margins for APIs will just get slimmer. Plus model capabilities are plateauing. Once model improvement slows down enough, seems to me like the battle is to be fought in the application layer. Whoever can make the killer app will capture the market.

Something worth noting is that ChatGPT currently is the killer app -- DeepSeek's current chart-topping app notwithstanding (not clear if viral blip or long-term trend).

For me ChatGPT was not that useful for work, the killer app was Cursor. It’ll be similar for other industries, it needs to be integrated directly in core business apps.

Re: Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

#104

Post by ern »

keithwhor wrote: post #93
mbowcut2 wrote: post #86

DeepSeek has demonstrated that there is no technical moat. Model training costs are plummeting, and the margins for APIs will just get slimmer. Plus model capabilities are plateauing. Once model improvement slows down enough, seems to me like the battle is to be fought in the application layer. Whoever can make the killer app will capture the market.

Something worth noting is that ChatGPT currently is the killer app -- DeepSeek's current chart-topping app notwithstanding (not clear if viral blip or long-term trend).

ChatGPT Plus gives me a limited number of o1 calls and o1 doesn't have web access, so I mostly have been using 4o in the last month and supplementing it with DeepSeek in the last week, for when I need advanced reasoning (with web search in DeepSeek as a bonus).

The killer app had better start giving better value, or I'd gladly pay the same amount of DeepSeek for unlimited access if they decided to charge.

Re: Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

#105

Post by tonyhart7 »

coffeebeqn wrote: post #76
tonyhart7 wrote: post #18

and microsoft literally spent 80 billions on top of its, like bro imagine 80 billions dollar company is like top 0,01 percent and that valuation would crumble because of deepseek

What did they spend it on? I’m sure the next gen models will still be better to train on that new infra

that total investment (so far), I believe Microsoft has back deal as well with open ai outside series funding investment

well since most of that money comeback anyway to MS since OAI use Azure heavily but it still a lot of money and stock value of OAI would tank sooner or later when competitor like deepseek come

Re: Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

#106

Post by oefnak »

SlightlyLeftPad wrote: post #63
handfuloflight wrote: post #55

Earlier quoted context omitted.

Oh yeah—tell that to my AI pipelines on my local compute.

Local pipelines are great and all for now but there’s practically no way those will be able to keep up with server based models long-term. If those are useful for you today, that’s great.

You're talking in a thread about Deepseek...

Re: Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

#107

Post by physicsguy »

azinman2 wrote: post #95
physicsguy wrote: post #85

Earlier quoted context omitted.

The interesting part is that distillations based on reinforcement learning based models are performing so well. That brings the cost down dramatically to do certain tasks.

I thought the distillations were SFT only?

They're SFT on the chain of thought output of R1

Re: Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

#108

Post by rasz »

agent281 wrote: post #45
manquer wrote: post #13

Earlier quoted context omitted.

I think it will have an adverse affect to funding ecosystem. The inevitable haircut all the funds are going to take in OpenAI and other AI startups when revenue fails to materialize[1] will herald a bust cycle and lot more circumspection in large investments like it happened few years back when a large number of Soft Bank investments did not pan out, notably most of them relied on big funding rounds to muscle out oth…

If it happens, I call dibs on the name "AI Nuclear Winter".

AI China Syndrome?

Re: Why OpenAI's $157B valuation misreads AI's future (Oct 2024)

#109

Post by impossiblefork »

drakenot wrote: post #72
Animats wrote: post #65

Earlier quoted context omitted.

Does anyone know how Deepseek does it yet?

(Summary from Reddit) - fp8 instead of fp32 precision training = 75% less memory - multi-token prediction to vastly speed up token output - Mixture of Experts (MoE) so that inference only uses parts of the model not the - entire model (~37B active at a time, not the entire 671B), increases efficiency - PTX (basically low-level assembly code) hacking in old Nvidia GPUs to pump out as much performance from their old H8…

They also use some kind of factorized attention that somehow leads to compression of tokens (I still haven't read their papers, so I can't be clearer than this).
Post reply on HN