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

New models and developer products

openai.com

Post reply on HN
Print view View original

331–340 of 561 posts

Re: New models and developer products

#331

Post by ianbicking »

modeless wrote: post #61
ianbicking wrote: post #37

Earlier quoted context omitted.

Do you also get those hallucinations just on silence? I kind of wonder if they had a bunch of training data of video with transcripts, but some of the video/audio was truncated and the transcript still said the last speech, and so now it thinks silence is just another way of signing off from a TV program. IMHO the bottleneck on voice now is all the infrastructure around it. How do you detect speech starting and stopp…

You're right, I think that's exactly what happened. Silence is when you get the most hallucinations. But there is a trick supported by some implementations that helps a lot. Whisper does have a special token that it predicts for silence. You can look at the probability of that token even when it's not picked during sampling. Hallucinations often have a relatively high probability for the nospeech token compared to ac…

This is what you mean? https://apps.microsoft.com/detail/9NC624PBFGB7

But I don't have Windows :(

Re: New models and developer products

#332

Post by caesil »

atleastoptimal wrote: post #324

Given that their main goal is still AGI, how does offering better developer tools and nifty custom models that can look at your dog for you help? Is it just bolstering revenue? They said they don't use API input to train their models so it isn't making them constantly smarter via more people using them.

AGI will be a system of different agents working together, not one mega-model.

Re: New models and developer products

#333

Post by Aeolun »

atleastoptimal wrote: post #324

Given that their main goal is still AGI, how does offering better developer tools and nifty custom models that can look at your dog for you help? Is it just bolstering revenue? They said they don't use API input to train their models so it isn't making them constantly smarter via more people using them.

Probably have more devs than they know what to do with at this point, so might as well spread them over the existing offerings while having the core work on AGI.

Re: New models and developer products

#335

Post by candiddevmike »

atleastoptimal wrote: post #324

Given that their main goal is still AGI, how does offering better developer tools and nifty custom models that can look at your dog for you help? Is it just bolstering revenue? They said they don't use API input to train their models so it isn't making them constantly smarter via more people using them.

They're in the AGI business the same way Tesla is in the self driving car business

Re: New models and developer products

#336

Post by dom96 »

Aeolun wrote: post #329
lolinder wrote: post #318

Earlier quoted context omitted.

> However, from a code perspective, it is a function of the key and a relatively small API. You're thinking of traditional apps and APIs. In an AI application, most of the work is in prompt engineering, not wiring up the API to your app. Prompts that work well for one model will fail horribly for another. People spend months refining their prompts before they're safe to share with users, and switching platforms will…

I’d be more worried about this if OpenAI had a track record of increasing prices, but the opposite happens. I get more for the same price basically every 6 months.

Sure, they are decreasing the prices right now. But once it comes time for them to become profitable they can easily reverse course.

Re: New models and developer products

#337

Post by passion__desire »

bkfh wrote: post #232

I wonder how many startups are obsolete after each OpenAI product release

Not related to your comment. But I see so much future in what ChatGPT can do.

Imagine giving a list of [Input Output] pairs, write a minimal program fitting the description in any language, even an Excel macro. Input, Outputs could in future be application interactions.

Adding onto it, imagine a future model where it understands shader toy scripts and its corresponding visual output.

This is like program fitting just as we have techniques for curve fitting and line fitting over a series of data points.

I am super pumped and excited for the future.

Re: New models and developer products

#338

Post by gwern »

leobg wrote: post #175
gwern wrote: post #103

> We’re also launching a feature to return the log probabilities for the most likely output tokens generated by GPT-4 Turbo and GPT-3.5 Turbo in the next few weeks, which will be useful for building features such as autocomplete in a search experience. This is very surprising to me. Are they not worried about people not just training on GPT-4 outputs to steal the model capabilities, but doing full blown logit knowled…

How many GBs worth of logits would you need to reverse engineer their model? Also, if it’s a conglomerate of models that they’re using, you’d end up in a blind alley.

Considering how well simply reusing GPT-3.5/4 outputs has worked to juice rival model performance, at least in relatively narrow benchmarking, I dunno how many GBs it'd take, but probably not that many, and it's a straightforward easy way to turn money into performance at a much lower cost than buying a few thousand more H100s.

Re: New models and developer products

#339

Post by JacobThreeThree »

somsak2 wrote: post #298
MagicMoonlight wrote: post #239

Earlier quoted context omitted.

Well it won’t because they’ll use the model you paid for and take your customers.

How do you square this with OpenAI's assertion that they never use data from enterprise customers for their own training? Are you suggesting they're lying?

They don't have to be currently lying for this to be a valid concern.

Clauses in terms of service are routinely updated or removed.

Re: New models and developer products

#340

Post by dragonwriter »

morkalork wrote: post #87
whytai wrote: post #71

Every day this video ages more and more poorly [1]. categories of startups that will be affected by these launches: - vectorDB startups -> don't need embeddings anymore - file processing startups -> don't need to process files anymore - fine tuning startups -> can fine tune directly from the platform now, with GPT4 fine tuning coming - cost reduction startups -> they literally lowered prices and increased rate limits…

If you want to be a start-up using AI, you have to be in another industry with access to data and a market that OpenAI/MS/Google can't or won't touch. Otherwise you end up eaten like above.

> a market that OpenAI/MS/Google can't or won't touch.

But also one that their terms of service, which are designed to exclude the markets that they can't or won't touch, don't make it impractical for you to service with their tools.

Post reply on HN