True Positive Weekly #103
The most important artificial intelligence and machine learning news and articles
Hey, in this issue: companies are failing to convince staff of AI benefits; pioneers of reinforcement learning win the Turing Award; LLMs in medicine; Gemma 3 - the most capable model you can run on a single GPU; Self-improving LLMs; the value of reinforcement learning in finetuning; and more.
[FT] Companies are failing to convince staff of AI benefits
[FT] Who needs revenue when you’re a multibillion-dollar AI start-up?
LLMs in medicine: evaluations, advances, and the future
[Models] Introducing Gemma 3: The most capable model you can run on a single GPU or TPU
[Model] QwQ-32B: Embracing the power of reinforcement learning
[ArXiv] LADDER: Self-improving LLMs through recursive problem decomposition
[ArXiv] All roads lead to likelihood: The value of reinforcement learning in fine-tuning
Enjoy the newsletter? Please help us make it bigger and better by sharing it with colleagues and friends.