Sam Altman and Ilya Sutskever, June-5-2023. «Could the open source LLM potentially match GPT-4's abilities… or am I wasting my time on 13B Wizard?» Ilya: «…there will always be a gap, and it may even increase over time» But he was right, you know. Whale bailed us out.

To be fair, releasing the models openly was the company’s (deepseek’s) decision. It’s a bit different from the traditional open-source paradigm. It was a major effort by teams of engineers managed by one company, not a cumulative effort. The engineering part of it, not the research.

Our hottest launch of 2024 is here! 🔥 Who will take up the challenge of aerial duels? 🛩️ "Warhammer 40,000: Dakka Squadron" is available now on the Nintendo eShop! More info:

Whats happening with safe superintelligence or whatever ilyas new company is called anyways? I remember a big deal about it but ive never seen any actual releases

They have said they're not going to release anything until ASI

do you feel transformer marginal benefit is in declining?

Transformer is fine. We haven't exhausted the limits of data improvement to really blame it

The absolute hubris of that response. "But only muh big corps can do AI". This also was during the time the US govt was telling all the other startups "dont even try" to compete with the approved player list. Turntables turned lmao.

All hail whale

Whale part 2 soon

Qwen 3 14B is GPT-4

probably not true in like ~9 months