• What a modern LLM does during training is, essentially, very very quickly skim the textbook, the words just flying by , not spending much brain power on it.


• Rather, when you or I read that math textbook, we read a couple pages slowly; then have an internal monologue about the material in our heads and talk about it with a few study-buddies; read an

... See more

Perhaps R1’s biggest breakthrough is the confirmation that you no longer need enormous data centers or thousands of labelers to push the limits of LLMs. If you can define what “correctness” means in your domain —whether it’s coding, finance, medical diagnostics, or creative writing— you can apply reasoning-oriented RL to train or fine-tune your own

... See more

研究人员发布了一篇新论文，表明机器人无需复杂的视频处理也能从YouTube学习技能。 这项研究充满前景且非常有趣。 https://t.co/21VqNRvwj2

core components of Deep RL that enabled success like AlphaGo: self-play and look-ahead planning.

Self-play is the idea that an agent can improve its gameplay by playing against slightly different versions of itself because it’ll progressively encounter more challenging situations. In the space of LLMs, it is almost certain that the largest portion o... See more

目前做大模型+机器人的 最大难点,在于获取数据 。谷歌和微软都选择了从收集家用动作入手,未来其他场景,比如办公室、街道、商场、工厂,如果能训练出来一个个小模型,融入到底层大模型里面,将会是创业公司的重要壁垒。

赛道特点?

1)劳动力替代/补强需求下的大市场(TAM),远期2C场景是万亿规模产业链;2)人工智能技术驱动下高进入壁垒(特别是偏具身模型的算法公司);3)存在规模效应带来头部集中(类比新能源车竞争格局);4)目前智能发展阶段仍然较早