Artificial General Intelligence (AGI) News and Discussions

[spryfusion]

SIMA 2: An Agent that Plays, Reasons, and Learns With You in Virtual 3D Worlds

Last year, we introduced SIMA (Scalable Instructable Multiworld Agent), a generalist AI that could follow basic instructions across a wide range of virtual environments. SIMA was a crucial first step in teaching AI to translate language into meaningful action in rich, 3D worlds.

Today we’re introducing SIMA 2, the next milestone in our research creating general and helpful AI agents. By integrating the advanced capabilities of our Gemini models, SIMA is evolving from an instruction-follower into an interactive gaming companion. Not only can SIMA 2 follow human-language instructions in virtual worlds, it can now also think about its goals, converse with users, and improve itself over time.

This is a significant step in the direction of Artificial General Intelligence (AGI), with important implications for the future of robotics and AI-embodiment in general.

Towards Scalable, Multitask Self-Improvement

One of SIMA 2’s most exciting new capabilities is its capacity for self-improvement. We’ve observed that, throughout the course of training, SIMA 2 agents can perform increasingly complex and new tasks, bootstrapped by trial-and-error and Gemini-based feedback.

For example, after initially learning from human demonstrations, SIMA 2 can transition to learning in new games exclusively through self-directed play, developing its skills in previously unseen worlds without additional human-generated data. In subsequent training, SIMA 2’s own experience data can then be used to train the next, even more capable version of the agent. We were even able to leverage SIMA 2’s capacity for self-improvement in newly created Genie environments – a major milestone toward training general agents across diverse, generated worlds.

[weatheriscool]

[wjfox]

[wjfox]

[weatheriscool]

[spryfusion]

[wjfox]

[spryfusion]

[weatheriscool]

Agi will be able to learn in real time, talk to you, reason and control a humanoid robot in human like ways.

Until something can do this it isn't agi. Human intelligence is a hell of a lot more complex then just finding answers from off the internet.

Data from star trek is my idea of agi.

[Cyber_Rebel]

I don't believe the "proto-AGI" which Demis Hassabis references needs to be embodied, but with the world model integration I imagine the robotics aspect of it would be much more impressive. Google actually has amazing robotics research in development which people might have forgotten about:https://deepmind.google/models/gemini-r ... -robotics/

This was making the head waves a few years back: https://deepmind.google/blog/rt-2-new-m ... to-action/

Non embodied AGI would be like Samantha from the Her movie or how EDI (Mass Effect 2) and the on-board Discovery AI from Star Trek were like. I agree that it needs to be able to learn online persistently.

It seems increasingly like the revolution which started from Google's famous "attention is all you need" paper might actually end with them too. That doesn't mean Open AI and Anthropic won't have AGI themselves in some capacity though. I doubt if Google announced Gemini being AGI in about 2-3 years from now just makes everyone else magically throw up their hands and stop developing their own.

[Yuli Ban]

[wjfox]

RE: Point 5 on that "fuzzy reality" list:

Sholto Douglas, Anthropic Researcher: Continual Learning will be solved in a satisfying way in 2026

Dario Amodei, CEO of Anthropic: We have evidence to suggest that continual learning is not as difficult as it seems

[Yuli Ban]

[wjfox]

Paper: https://test-time-training.github.io/e2e.pdf

[wjfox]

AI Futures Model: Dec 2025 Update

https://blog.ai-futures.org/p/ai-future ... 025-update

"At the time the Full Takeoff Model was created it predicted AGI in 2040, with the parameter settings chosen by Davidson. But both compute and algorithmic progress has been faster than they expected. When the FTM is updated to take into account this new data, it gives shorter medians in the late 2020s or early 2030s. Meanwhile, with GATE’s median parameters, it predicts AGI in 2034."

[weatheriscool]

[firestar464]

[firestar464]

Honestly, I do think that motor intelligence should be factored in, because the benchmark is people.

Key here is "average human."

[spryfusion]

[wjfox]