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

Introducing ChatGPT and Whisper APIs

openai.com

Post reply on HN
Print view View original

211–220 of 696 posts

Re: Introducing ChatGPT and Whisper APIs

#211

Post by ChickenNugger »

goldfeld wrote: post #129
ChickenNugger wrote: post #58

Earlier quoted context omitted.

For real! This stuff is moving fast . It feels like just last week I was posting about how it's going to change...art. And now there are hilarious deepfake memes of past and current presidents shit talking about video games. There are a handful of ML art subs that have pretty amazing stuff daily. Especially the NSFW ones, which if you've studied any history of media VHS/DVD/Blu-ray/the internet, porn is a major innov…

What are some niche ML art subs to hang around in? Excepting the NSFW..

I don't know about niche, but MachineLearning and StableDiffusion are the only SFW ones.

FWIW, the NSFW ones are unstable_diffusion, sdforall, sdnsfw, aipornhub

Re: Introducing ChatGPT and Whisper APIs

#212

Post by cypress66 »

anonylizard wrote: post #42
krelian wrote: post #33

Earlier quoted context omitted.

I thought it was Midjourney who stole their thunder. Stable Diffusion is free but it's much harder to get good results with it. Midjourney on the other hand spits out art with a very satisfying style.

You are like 2 months out of date. Stable diffusion now has a massive ecosystem around it (civitai/automatic1111), that when used well, completely crushes any competitors in terms of the images it produces. Midjourney is still competitive, but mostly because its easier to use. Dalle2 will get you laughed out of the room in any ai art discussion.

Don't forget controlnet, which is a game changer.

Re: Introducing ChatGPT and Whisper APIs

#213

Post by gremlinsinc »

Cantinflas wrote: post #53
anonylizard wrote: post #42

Earlier quoted context omitted.

You are like 2 months out of date. Stable diffusion now has a massive ecosystem around it (civitai/automatic1111), that when used well, completely crushes any competitors in terms of the images it produces. Midjourney is still competitive, but mostly because its easier to use. Dalle2 will get you laughed out of the room in any ai art discussion.

It's amazing that "being two months out of date" in AI means that you are already a dinosaur

That's what the singularity is all about, a moment in time when 2 seconds late turns you into a dinosaur, be greatful it's 2 months, not 2 weeks, 2 days, or 2 minutes.

Re: Introducing ChatGPT and Whisper APIs

#214

Post by thewataccount »

minimaxir wrote: post #138
vishal0123 wrote: post #82

Earlier quoted context omitted.

> I have no idea how OpenAI can make money on this. I did some quick calculation. We know the number of floating point operations per token for inference is approximately twice the number of parameters(175B). Assuming they use 16 bit floating point, and have 50% of peak efficiency, A100 could do 300 trillion flop/s(peak 624[0]). 1 hour of A100 gives openAI $0.002/ktok * (300,000/175/2/1000)ktok/sec * 3600=$6.1 back.…

It's speculated that ChatGPT uses 8x A100s, which flips the conclusion. Although the ChatGPT optimizations done to reduce costs could have also reduced the number of GPUs needed to run it.

Wait 8x total? For everyone at once?

Re: Introducing ChatGPT and Whisper APIs

#215

Post by tstrimple »

jejeyyy77 wrote: post #37
tstrimple wrote: post #27

Earlier quoted context omitted.

Bound to happen, so establish yourself as deeply as possible as quickly as possible. Once folks are hooked up to these APIs, there's a cost and friction to switching. This just feels like a land grab that OpenAI is trying to take advantage of by moving quickly.

Is there though? It's just a matter of swapping out $BASE_API_URL.

Most of the clients I'm working with aren't interested in the base level of service. They are looking to further train the models for their specific use cases. That's a much higher barrier to switch than replacing an API. You've got to understand how the underlying models are handling and building context. This sort of customer is paying far more than the advertised token rates and are locked in more tightly.

Re: Introducing ChatGPT and Whisper APIs

#216

Post by osigurdson »

minimaxir wrote: post #2

> It is priced at $0.002 per 1k tokens, which is 10x cheaper than our existing GPT-3.5 models. This is a massive, massive deal. For context, the reason GPT-3 apps took off over the past few months before ChatGPT went viral is because a) text-davinci-003 was released and was a significant performance increase and b) the cost was cut from $0.06/1k tokens to $0.02/1k tokens, which made consumer applications feasible wit…

>> may be better than in-house finetuned LLMs

I don't think this competes with fine-tuned models. One advantage of a fine tune is it makes use of your own data.

Re: Introducing ChatGPT and Whisper APIs

#217

Post by osigurdson »

vishal0123 wrote: post #82
minimaxir wrote: post #2

> It is priced at $0.002 per 1k tokens, which is 10x cheaper than our existing GPT-3.5 models. This is a massive, massive deal. For context, the reason GPT-3 apps took off over the past few months before ChatGPT went viral is because a) text-davinci-003 was released and was a significant performance increase and b) the cost was cut from $0.06/1k tokens to $0.02/1k tokens, which made consumer applications feasible wit…

> I have no idea how OpenAI can make money on this. I did some quick calculation. We know the number of floating point operations per token for inference is approximately twice the number of parameters(175B). Assuming they use 16 bit floating point, and have 50% of peak efficiency, A100 could do 300 trillion flop/s(peak 624[0]). 1 hour of A100 gives openAI $0.002/ktok * (300,000/175/2/1000)ktok/sec * 3600=$6.1 back.…

This would be a really fun optimization challenge for sure!

Re: Introducing ChatGPT and Whisper APIs

#218

Post by valine »

osigurdson wrote: post #210

I always wondered if ChatGPT was somehow stateful. I assumed that it was not and the statefulness was simulated. Assumption validated.

I don’t believe this was ever in question. You can think of the model as a giant function that takes a list vectors as input and spits out a new vector. If you want the model to remember something, you have to include it in the list of input vectors for every request going forward.

Re: Introducing ChatGPT and Whisper APIs

#219

Post by sebzim4500 »

visarga wrote: post #161
minimaxir wrote: post #2

> It is priced at $0.002 per 1k tokens, which is 10x cheaper than our existing GPT-3.5 models. This is a massive, massive deal. For context, the reason GPT-3 apps took off over the past few months before ChatGPT went viral is because a) text-davinci-003 was released and was a significant performance increase and b) the cost was cut from $0.06/1k tokens to $0.02/1k tokens, which made consumer applications feasible wit…

They probably shrunk the model from 175B to 17B. That's your 10:1 price reduction.

Wouldn't that almost certainly lead to measurable loss of capabilities?

Re: Introducing ChatGPT and Whisper APIs

#220

Post by Kiro »

phas0ruk wrote: post #122

Feels like people running small websites monetised with ads will get killed. Why go to a recipe website to search for healthy meals for your kids if you can ask chat gpt in Instacart? :(

These small websites have been dead for years and replaced with SEO spam.
Post reply on HN