michael abbott

so @mabb0tt and I are once again volunteering to teach cs153.stanford.edu there are so many new frontiers to be pioneered thank you to our speakers like @karpathy @bhorowitz @brendaniribe @DavidBaszucki @LiamFedus @ekindogus @sama for investing in the next generation

so @mabb0tt and I are once again volunteering to teach cs153.stanford.edu there are so many new frontiers to be pioneered thank you to our speakers like @karpathy @bhorowitz @brendaniribe @DavidBaszucki @LiamFedus @ekindogus @sama for investing in the next generation

so @mabb0tt and I are once again volunteering to teach cs153.stanford.edu there are so many new frontiers to be pioneered thank you to our speakers like @karpathy @bhorowitz @brendaniribe @DavidBaszucki @LiamFedus @ekindogus @sama for investing in the next generation

so @mabb0tt and I are once again volunteering to teach cs153.stanford.edu there are so many new frontiers to be pioneered thank you to our speakers like @karpathy @bhorowitz @brendaniribe @DavidBaszucki @LiamFedus @ekindogus @sama for investing in the next generation

so @mabb0tt and I are once again volunteering to teach cs153.stanford.edu there are so many new frontiers to be pioneered thank you to our speakers like @karpathy @bhorowitz @brendaniribe @DavidBaszucki @LiamFedus @ekindogus @sama for investing in the next generation

In Marin, we are trying to get really good at scaling laws. We have trained models up to 1e22 FLOPs and have made a prediction of the loss at 1e23 FLOPs, which @WilliamBarrHeld is running. This prediction is preregistered on GitHub, so we'll see in a few days how accurate our prediction was. What we want is not just a single model but a training recipe that scales reliably.

I partly agree. There will always be top talent or perceived top talent. Perhaps it would have been better to focus on the “regular” $1M++ p.a. packages of many AI technical staff at top labs. What are the issues there? It’s partly that the human cost is small vs. the hardware cost. But I think a big part of it is the restricted supply of people with LLM expertise and experience, in considerable part because universities have not had sufficient resources for research students to be able to become skilled in larger scale neural model building. We need many more people able to train at least small language models in academia or other open contexts, like: iclr.cc/virtual/2025/i… marin.community

Congratulations to senior @olivialabbott, who scored 14 goals and tallied 10 assists across 4 days, while competing for the USA Lacrosse U19 team in the Rivalry Challenge! 🔗 usalacrosse.com/magazine/natio…