Hacker News Reader

Comment by Imnimo | Hacker News Reader

I know everyone's on text-embedding-ada-002, so these particular embedding deprecations don't really matter, but I feel like if I were using embeddings at scale, the possibility that I would one day lose access to my embedding model would terrify me. You'd have to pay to re-embed your entire knowledge base.

View on HN

They said in the post,

> We recognize this is a significant change for developers using those older models. Winding down these models is not a decision we are making lightly. We will cover the financial cost of users re-embedding content with these new models. We will be in touch with impacted users over the coming days.

That's what I always thought. Someday they will come up with a new embedding model, right?

What I don’t understand is why is an API needed to create embeddings. Isn’t this something that could be done locally?

You would need to have a local copy of the GPT model, which are not exactly OpenAI's plans.

For embeddings, you can use smaller transformers/llms or sentence2vec and often get good enough results.

You don't need very large models to generate usable embeddings.

You are correct, I assumed parent was referring to specific embeddings generated by OpenAI LLMs.

It’s cheaper to use OpenAI. If you have your own compute, sentence-transformers is just as good for most use cases.

Yes. The best public embedding model is decent, but I expect it’s objectively worse than the best model from OpenAI.

Sure, but I don't know of any models you can get local access to that work nearly as well.

If you read the article they state they will cover the cost of re-embedding your existing embeddings.

Reply on news.ycombinator.com