Ahh I fell for it. I was fairly convinced by the 100+ 40GB files as weights to download, but when I saw the paper was written in a word doc I remembered to check the date.

One interesting thing to note is that I could tell they used ChatGPT rather than GPT-4 to do the fake writeup. ChatGPT likes to generate paragraphs shown in their paper, GPT-4 will do full latex commands, inline citations, and even (a made up) bibtex for the citations.

GPTrollion, more like it. It's an April's fools joke, not a very good one I'd say.

Also submitted earlier here:

https://news.ycombinator.com/item?id=35402147

This is an April fools joke.

From the linked paper: > It is important to note that the concept of GPTrillion and its capabilities are entirely fictional and were created as an April Fools joke for the AI community. GPTrillion does not actually exist, and the claims made in this paper are not based on any real experiments or data.

Please flag this off the front page.

So everyone is basically totally ignoring the open letter calling for pause in large AI experiments?

So nobody else wastes time excitedly reading the paper like I did:

>In conclusion, while GPTrillion is a fictional model, this paper can still serve as a humorous commentary on the state of language modeling and the challenges that come with training and developing large models. However, it is essential to maintain scientific integrity and ensure that claims made in research papers are based on real experiments and data.

Not sure I get the joke, but it did manage to fool me into getting my hopes up briefly.