Twitxy Saved

Jeff Clune

Verified account This account is verified. Learn more

@jeffclune

Co-founder, Recursive. Professor, CS, U. British Columbia. CIFAR AI Chair, Vector Institute. | ML, AI, deep RL, deep learning, AI-Generating Algorithms (AI-GAs)
Paradise (Vancouver, BC) jeffclune.com Joined July 2010
448 Following 34K Followers 4.4K Posts

Latest posts 5

Pinned post

Jeff Clune Verified account @jeffclune
Thrilled to share that we founded Recursive to create AI that safely conducts experiments on how to improve itself in an open-ended process of endless, automated scientific discovery. As I wrote in my 2019 AI-generating algorithms paper, this will likely be the fastest path to superintelligence. Our work since has shown the power of this approach. Excited to scale up and improve upon ideas like the Darwin Gödel Machine, HyperAgents, ADAS, OMNI, ALMA, The AI Scientist, PromptBreeder, Rainbow Teaming, Automated Capability Discovery, and other work on open-ended and AI-generating algorithms. We’ve assembled a dream team of researchers and significant resources to pursue this vision. My amazing co-founders are pictured here, and we have an all-star team of founding members (we’re over 25 and growing).

Please join us if you are interested! Follow our progress @Recursive_SI
56 replies 43 reposts 636 likes 124K views View on X
Jeff Clune Verified account @jeffclune
Amazing and rewarding to see our small-team research effort (led by @ashrewards , @jparkerholder, @_rockt et al.) take on a life of its own as an ever-improving flagship Google product! For me personally, a crazy idea I proposed in my 2019 AI-GA paper (a NN as the entire environment including pixels, sound, etc.) barely worked in Genie1 (2024), worked decently in Genie 2 (8 months later), became incredible in Genie3 (9 months later), and now Genie4 seems ready for Hollywood and VR. What an arc to observe in such a short time. Wild times! Now imagine open-ended learning in Genie. ✨ 🚀
Google DeepMind @GoogleDeepMind Oct 27
This is Genie 4, our most realistic and interactive world model yet. 🌐

Explore dynamic environments in real time to see how the scene responds, sounds, and how the people in it move, react, and speak.
View quoted post on X
10 replies 3 reposts 100 likes 8.9K views View on X
Jeff Clune Verified account @jeffclune
Scary!
Cohere Verified account @cohere Oct 26
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.

Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.

Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.

But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?

We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.

Read the full report: cohere.com/blog/concernin…
View quoted post on X
3 replies 1 reposts 20 likes 5.2K views View on X