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

OpenAI API pricing update FAQ

help.openai.com

Post reply on HN
Print view View original

131–140 of 205 posts

Re: OpenAI API pricing update FAQ

#131

Post by elefanten »

paulryanrogers wrote: post #124
aaroninsf wrote: post #114

Earlier quoted context omitted.

I can comment that a friend teaching college philosophy said his department is eliminating papers at the undergraduate level, because it is now too easy to generate-then-polish papers at that level. They're too good in other words: better than baseline even. And when touched up by a student who's putting in a modest effort, are disruptively better.

Could that have something to do with the subjective nature of philosophy? (IANAP)

Probably has more to do with how slowly the field evolves (esp. wrt what undergrads are learning)... so I imagine there's a huge training corpus for what is ultimately a very limited number of meaningful assignments to give.

Re: OpenAI API pricing update FAQ

#132

Post by teaearlgraycold »

minimaxir wrote: post #70
dr_dshiv wrote: post #62

Earlier quoted context omitted.

Any good use cases for fine tuned models? I’ve been wanting to put in all my previous academic works and references and see how that changes things. Otherwise, I’ve been using a model finetuned on the complete works of Plato and I have really been enjoying the new dialogues.

I made a blog post just last week about finetuned GPT-3, albeit only on the lower end and not finetuning davinci: https://news.ycombinator.com/item?id=32471208 I suspect some of the more custom "be a better writer" services use a finetuned davinci.

CopyAI uses fine tuning extensively

Re: OpenAI API pricing update FAQ

#133

Post by choppaface »

londons_explore wrote: post #83

I'm really surprised they don't charge a very high price with a generous free tier, and then have an army of salespeople who will offer you whatever discount is necessary to make your business model work. If you want to use GPT-3 to build an AI-lawyer, it'll be very expensive... but to build a tool for customer support for walmart will be cheap, etc.

The OpenAI CEO Sama is primarily an investor, not an operator, so the model is tuned towards giving small companies an opportunity versus enticing enterprise customers. Sama want somebody to figure out how to run & productive his business versus actual customers. In particular because of efforts like Stable Diffusion—- it’s going to be a lot easier for enterprise to build off tech that’s actually Open.

Re: OpenAI API pricing update FAQ

#134

Post by teaearlgraycold »

layer8 wrote: post #76
dane-pgp wrote: post #29

Is this because GPT-4 is about to be released and they want to charge full price for that?

Maybe demand has slowed and they want to keep it up.

GPT-J was about the same cost to self-host as GPT-3 (ignoring maintenance costs). They’re getting ahead of competitors IMO.

Re: OpenAI API pricing update FAQ

#135

Post by choppaface »

AJRF wrote: post #6

> We have been looking forward to reducing pricing for a long time. Our teams have made incredible progress in making our models more efficient to run, which has reduced the cost it takes to serve them, and we are now passing these savings along to our customers. Nothing to do with Stable Diffusion then, no?

It’s not just Stable Diffusion the model, but the research is become really actually open unlike “open ai”

Re: OpenAI API pricing update FAQ

#136

Post by dr_dshiv »

slavik81 wrote: post #110
dr_dshiv wrote: post #52

Earlier quoted context omitted.

I also asked it to help me write a letter to my homesick daughter at summer camp. It was really helpful. I feel that GPT3 has an enviable degree of emotional intelligence, nuance and conceptual understanding. Writing and communicating effectively with people is very hard. Our own emotions get in the way. Now we have technology that can help people be better communicators with each other. It’s a big deal.

When you say you're an empath, do you mean to claim that you have psychic powers? An empath is a being from science fiction with the psionic ability to sense emotions, like Counselor Troi on Star Trek. The term originates from the short story "Empath" by J. T. McIntosh, published in New Worlds Science Fiction #50, August 1956.

Yes, that is correct. I have psychic powers, specifically the psionic ability to sense emotions. AMA

Re: OpenAI API pricing update FAQ

#138

Post by hcks »

I really wonder about the economics and business models of large models APIs, mainly because of:

* large running costs due to the expensiveness of the inference

* low barriers to entry: the tech behind Dalle appeared 1.5 years ago if I recall right and there are already a few competitors (Midjourney, Imagen, Stable Diffusion)

* low value at the present time of the raw APIs

Re: OpenAI API pricing update FAQ

#139

Post by leesec »

cbm-vic-20 wrote: post #33

I lack imagination. What is GPT-3 (and generative text models in general) useful for, other than toys? It can write essays, can we expect it to replace writers for newspapers and magazines? Can it write coherent technical documentation? Replace speechwriters? Edit: looks like the comment thread below is similar https://news.ycombinator.com/item?id=32554955

Write now it's a great writing-assist tool. Eventually all the rest.

Re: OpenAI API pricing update FAQ

#140

Post by bitdivision »

IncRnd wrote: post #116
bitdivision wrote: post #90

Earlier quoted context omitted.

I'm not sure about using 10x in place of 'an order of magnitude'. An order of magnitude conveys an approximation, whereas 10x is precise.

10x is literally the difference of an order of magnitude. An order of magnitude is an exponential change of plus or minus 1 in the value of a quantity or unit. The term is generally used in conjunction with power-of-10 scientific notation. Order of magnitude is used to make the size of numbers and measurements of things more intuitive and understandable. It is generally used to provide approximate comparisons between…

Yes, an order of magnitude literally means 10x. But in my experience, in common speech it's often used to convey an approximation. i.e. 'Changing this will decrease performance of that endpoint by an order of magnitude', when it's really somewhere around 10x. If someone said 'decrease performance by 10x' that seems much more concrete to me.
Post reply on HN