Loading activity…
@ianosband
Research scientist at DeepMind (ex OpenAI, Stanford) working on decision making under uncertainty.
AI
A user argues that labs willing to bootstrap or distill build great models faster than labs building self-sufficient stacks.
AI
In an X thread, Ian Osband argued that policy gradient can be exact in image classification and still perform far worse than cross-entropy, citing his own reported 4% versus 62% ImageNet comparison and proposing a planning-style “horizon loss.”
AI
Schmidhuber cites a 1993 paper on a self-referential weight matrix and a 2022 paper on a matrix that learns to modify itself.
AI
Cognitive scientist and LLM critic made the remark in a social post.
