Isn't it perfectly obvious that this did not happen?

PaLM 2 Unicorn variant claims to be GPT-4 competitive on some benchmarks in the paper. That already should resolve it NO.

Google Gemini is currently training, is intended to beat GPT-4, and is being trained with "5x as many FLOPs as GPT-4" https://www.semianalysis.com/p/google-gemini-eats-the-world-gemini

Google Gemini Eats The World – Gemini Smashes GPT-4 By 5X, The GPU-Poors

Compute Resources That Make Everyone Look GPU-Poor

Google has woken up, and they are iterating on a pace that will smash GPT-4 total pre-training FLOPS by 5x before the end of the year. The path is clear to 20x by the end of next year given their current infrastructure buildout.

no way unfortunately

https://www.theverge.com/2023/3/30/23662101/ftc-openai-investigation-request-caidp-gpt-text-generation-bias

@connorwilliams97 Even if this goes through, no way will it carry over to other countries.

Probability 15% IMO. The letter itself won't, but it could be the catalyst of something bigger.

based and you-pilled

@metachirality Probably fake

@metachirality yeah fake

@metachirality Yeah, looks like it was deleted

lol

The main way I could see this occurring in the next year is if one argues that OpenAI has a substantial lead such that they are the only group that can train an "AI system more powerful than GPT-4" at the moment and that they decide to pause for 6 months. That seems extremely unlikely given (i) their view that the safest quadrant for AGI is short timelines and slow takeoff, and (ii) the majority view in AI safety that we're better off having one company far ahead of others that they can take time to slow down at the end.

@JacyAnthis they're basically not ahead at all

@L What do you see as the groups who are even with them? (or if not even, how far behind them do you think the next runner up is)