i think everyone in ai should think about what their "line" is.
where if your company knowingly crosses that line and won't walk it back, you'll walk away
this line is personal, will be different for everyone, and can feel far-fetched even. you don't have to share it with anyone, but i recommend writing it down as an anchor for your future self
inspired by two people i deeply respect who just did, from different labs
This patent brings together ideas from:
Trustworthy AI
Behavioral modeling
Swarm systems
And was such a rewarding collaboration with VJ, Selvi Arumugam & Badri.
Why does this matter?
Because device are becoming targets. Which means a large number of humans could be impacted without knowing what hit them.
This gives us a new layer of defense: asking not just who a system claims to be, but how it behaves when probed.
Really sad to hear that Papers with Code is shutting down.
I still remember stumbling onto it years ago when the ML community felt like a small village of nerds sharing arxiv/openreview links.
Thankfully, Hugging Face has stepped in with an alternative page for trending papers: https://t.co/Agvls8mi5S.
I hope it carries forward that spirit of discovery and open collaboration.
This year was a major paradigm shift, where we can solve problems end to end in natural language. With novel reinforcement learning techniques, we are able to train an advanced Gemini model on multi-step reasoning proof data, which advances the model's capabilities in terms of long-answer generation. Combined with the power of Deep Think model, where we enable the model to think more in parallel and longer, in the case of this year IMO 4.5 hours, we can solve P1, P2, P3, P4, P5 flawlessly!
Jokes aside, I spend a lot of time either editing code written by Claude, or telling it very specifically what method it needs to change.
Curious about all the stories about non-technical folks using AI models to write 100% of their code - myth or fact?