Hacker News Reader

Comment by ldjkfkdsjnv | Hacker News Reader

Goodbye all the startups trying to build LLMs/fine tuning/agents. OpenAI is so dominant all of these startups are solving sub problems of the generalized algorithms of OpenAI. Good for them for crushing the competition, finally a competent company broke the monopolies of Google/FB. All those highly paid researchers at Google/FB cannot release models that are remotely close. The emperor has no clothes

View on HN

How does this actually kill fine tuning? I don't think OpenAI can kill open models. People are not going to to stop wanting to keep their data in-house or at least more under their own control. And there are massive companies behind the open models.

they just announced custom models for enterprise

That will put a dent in it for sure. But still I feel like fine tuning is the most resilient aspect of it for open models.

OK but their secret sauce is out there. There are other foundational models. Curious to see if they can create lock-in or network effects with their APIs and such

isnt this the basis of the conspiracy theory like ideas around asking for AI regulation from the goverment

Realistically nothing performs as well as GPT-4 at the moment. We've seen others make advances but it's still best in class.

at the same time it's very hard to monetize GPT4 because the cost are so high

my uneducated guess is they may aim to get major share of income via selling ChatGPT Enterprise to Fortune 500 list, where cost of running probably is not much of a concern

There's thousands of niches where you have to really know the domain, which OpenAI will never bother doing. If startup is smart and has a moat, it's going to be OK.

The irony of this all is that ai has successfully replaced ai workers, and will continue doing so.

Reply on news.ycombinator.com