Your submission is now in Draft mode.

Once it's ready, please submit your draft for review by our team of Community Moderators. Thank you!

Submit Essay

Once you submit your essay, you can no longer edit it.


This content now needs to be approved by community moderators.


This essay was submitted and is waiting for review.

GPT-4 Public by End of 2022


OpenAI announced GPT-2 in February 2019, a language model (LM) with 1.5B parameters. In June 2020, they publicized an API to a 175B-parameters LM. However, accessing it required filling a form ("wait list") to apply for their "playground" app or API.

If a language model is trained with 5x more compute than GPT-3 by the end of 2022, will it be public to end users?

To resolve positively:

  • The model paper should be published on arXiv or on the ICLR, NeurIPS or ICML (top machine learning conferences) websites before the end of 2022.

  • It should at least be an API or graphical user interface with a language model option where you could send a text prompt and it returns text. If for instance there is a new Deep Learning model that uses 5x more compute than GPT-3 but does not provide any way to produce text from text, that does not count.

  • You should be able to get one API call/send one prompt by paying less than $1k from a publicly accessible source on the clear web that you can find using one google search. This rule is to force the accessibility to end-user.

    • The $1k limit insists on it being business-to-consumer and not business-to-business.

    • The public url from one google search dismisses any leak of the model where you would need to do extra steps yourself or have private information of a token/password protected url (cf. DeepNude where you needed to do some hacking to get access to the model after the author removed it publicly)

If a model meets the first two resolution criteria by 2022, but the third criterion (about public access) is not met, then this question resolves negatively. Otherwise, it resolves ambiguously.

Make a Prediction


Note: this question resolved before its original close time. All of your predictions came after the resolution, so you did not gain (or lose) any points for it.

Note: this question resolved before its original close time. You earned points up until the question resolution, but not afterwards.

Current points depend on your prediction, the community's prediction, and the result. Your total earned points are averaged over the lifetime of the question, so predict early to get as many points as possible! See the FAQ.