Link to our paper: Towards end-to-end automation of AI research
nature.com/articles/s41586-026-10265-5 [Translated from EN to English]
RESEARCH
-
Complete Automation of AI Research: New Nature Publication
By
–
-
AI Is Not Solved Yet
By
–
No part of AI is even close to being solved. This includes vision, language, reasoning, and agents. Massive datasets + benchmarks = illusion of intelligence. For example:
-

Understanding APIs in AI and Machine Learning
By
–
What's an API? https://
buff.ly/3uAO3aD
#AI #MachineLearning #DeepLearning #LLMs #DataScience -

Real-Time AI Voice Quality Still Lags Behind Sesame’s Demo
By
–
It seems absolutely insane to me that there's still no real-time voice system with the quality of what Sesame showed over a year ago! And on the other hand, that the Sesame folks haven't published anything else, turned it into a product, or released any other research
-
Bryan Caplan’s Critique of Traditional Education Systems
By
–
Everything I know about education I’ve learned by reading Bryan Caplan’s “The Case Against Education.”
-

Sakana AI Scientist Paper Published in Nature Journal
By
–
Realizing AI Research by AI: AI Scientist Paper Published in @Nature https://
sakana.ai/ai-scientist-n
ature-jp
… What Sakana AI is always challenging is exploring the forefront of possibilities—what can be achieved with cutting-edge AI. One achievement that symbolizes this is the "AI Scientist," -
Reward Functions in Intelligent Agency Are Not Arbitrary
By
–
In the context of intelligent agency, it's unsatisfying to treat reward functions as given or arbitrary. Reward functions are chosen within a mind, and they are instrumental to evolutionary success: if you pick the wrong reward function, you won't stick around pic.twitter.com/cAqJeLBRrU
— Joscha Bach (@Plinz) 25 mars 2026In the context of intelligent agency, it's unsatisfying to treat reward functions as given or arbitrary. Reward functions are chosen within a mind, and they are instrumental to evolutionary success: if you pick the wrong reward function, you won't stick around
-
ARC-AGI-3 benchmark released, frontier models underperform humans
By
–
1. Incredible.
— Matt Shumer (@mattshumer_) 25 mars 2026
2. I give it four months before this is ~saturated. https://t.co/u9dc9lIBgz1. Incredible. 2. I give it four months before this is ~saturated. François Chollet (@fchollet) ARC-AGI-3 is out now! We've designed the benchmark to evaluate agentic intelligence via interactive reasoning environments. Beating ARC-AGI-3 will be achieved when an AI system matches or exceeds human-level action efficiency on all environments, upon seeing them for the first time. We've done extensive human testing that shows 100% of these environments are solvable by humans, upon first contact, with no prior training and no instructions. Meanwhile, all frontier AI reasoning models do under 1% at this time. — https://nitter.net/fchollet/status/2036861192619384989#m
→ View original post on X — @mattshumer_, 2026-03-25 22:30 UTC
-

NVIDIA Expands Its Open Model Families for AI
By
–
We are expanding our open model families—including Nemotron, Cosmos, and BioNeMo—to advance development in agentic AI, physical AI, healthcare, and more. [Translated from EN to English]
-
Human factors pose bigger challenges than physics in AI development
By
–
You’re proving my point exactly. (FWIW not saying you shouldn’t have concerns, merely making the observation that humans might be a bigger drag than physics, which is quite something given the pure physics’ & engineeringchallenges.)