Ask HN: Anyone else see GPT-4 access as unfair?
GPT-3.5 and GPT-4 are worlds apart and having access to GPT-4 (API) is a competitive advantage.
Like countless others, I've been on the waitlist forever, and seeing other companies get access to GPT-4 means my project is on hold while they're moving ahead.
Am I the only one who sees this as an unfair practice by OpenAI?
47 comments
[ 2.7 ms ] story [ 92.5 ms ] threadThere is no different signup or waitlist for the api. It's just much more expensive to use gpt4 in favor of gpt-3.5-turbo.
The first example you find there for creating a chat-app let's you freely choose between the models.
https://platform.openai.com/docs/guides/chat
Just put the creditcard details in and do not forget to set usage-limits to your liking in the settings.
This way you do not have to fear higher costs than you are willing to accept. There is a standard cap of 120$ per month. That is the absolute maximum, if you do not contact them and raise it. I turned mine to 20$.
https://openai.com/pricing#language-models
I just hammered against it and found out you are totally right. Which is a real pain, because i have lots of things to test and checkout with more tokens.
I am working with a friend on something, and i know he has access, we share the code and tomorrow i wanted to start using GPT-4 with it.
So i am now as disappointed as you are.
Sorry for my confusion, but like i said, i had no prior need for a paid account (api) and the docs do nowhere mention that.
Also, does anyone know if Copilot is using GPT-4? I assume it is.
I've compared the same prompts between both, and GPT-4 is always better, except when you need speed or it's a very simple task.
For complex tasks, especially requiring some sort or social understanding (I'm building a group AI sharing tool) it has a capacity for nuance which is sometimes surprisingly good. ChatGPT is usually more straightforward.
If anyone hasn't played with gpt4 yet I'm looking for testers on my platform and I have access.
https://openai.com/research/gpt-4
Sounds like something they would do.
> Our mission is to ensure that artificial general intelligence benefits all of humanity.
Why is total fairness a prerequisite for benefiting humanity?
FWIW, I don't think that they're giving access based on that mission, but I don't think your claim has much merit. There is a possibility that unfairness and their mission are perfectly in sync.
What if you had to give away 10% of your company to get access to AWS (or any cloud provider)?
“Missions” are pure marketing. The only mission of a corporation is to make money.
I don’t think “fair” is even part of the idea.
Just goes against their mission, if that’s still a thing.
It's disappointing since Sama was a part of YC which does seem to care about merit and "fairness." So you'd hope he'd support the same principals to "support the little guy" and give everyone a shot at having the best tools.
At least GPT4 isn't good enough to use something like AutoGPT to actually be useful at self-developing large portions of software. Once AI is that good, API access will be critical. Entire infrastructure and first-mover advantage could be dependent on API access at that time... there would be AI pipelines that wouldn't have been possible months ago that a highly intelligent AI model could see to use instantly and build out in a day. Imagine if your competitor has access and you don't in that AI future.
I believe in the near future they will put AIs locally in your toaster and all toys, because they will become so cheap to embed offline.
Once the diminishing returns of improvements drop off or when it gets "good enough", then it won't matter.
> At least GPT4 isn't good enough to use something like AutoGPT
I worry that GPT-4 will continue to get better while access will still be limited.