What if an LLM could update its own weights?
Meet SEAL🦭: a framework where LLMs generate their own training data (self-edits) to update their weights in response to new inputs.
Self-editing is learned via RL, using the updated model’s downstream performance as reward.
Church lab automates 12 years of research in 2 days
Sometimes I have the privilege of receiving emails from people telling me about their breakthroughs. In such cases, it gives me great pleasure to help advance scientific knowledge by giving it a little reach with my modest
Time‑resolved cryo‑EM now films proteins every 5µs:
⚡️Lasers melt‑revitrify + 💧droplet mix‑and‑spray chips need just 100nL per shot, catching fleeting drug‑ready states 👀👀👀
In Science, researchers present a new method to safely and preferentially generate CAR T cells directly inside the body using targeted lipid nanoparticles that deliver mRNA directly to T cells.
The approach showed rapid and sustained immune reprogramming in preclinical models,
Fine. Then UK taxpayers working for companies you dont invest in, shouldn’t be subsiding pensions.
Aviva boss warns ministers over making savers invest in UK assets. thetimes.co.uk/article/5b9c8a…
Today's big biotech win is that we might be on the verge of a cure for type-1 diabetes🧵
Twelve diabetics were injected with stem cell-derived pancreatic islets.
They started producing insulin again.
One year in, 10/12 participants no longer needed to inject insulin.