#agi

Bloomberg reports DeepSeek is pushing forward with a $10.29 billion financing round. Founder Liang Wenfeng publicly reaffirmed commitment to open-source AI development and AGI over short-term commercialization.

A 2024 paper claiming to prove — via complexity theory — that machine learning cannot achieve human-level performance has been shown to be irreparably broken. The rebuttal paper is now published in the same journal, Computational Brain & Behavior.

A DeepMind employee publicly challenged private AI labs claiming to be on the path to AGI, arguing they should either go public or raise capital that ordinary people can invest in — otherwise, they're simply enriching billionaires.

Jack Clark, Anthropic co-founder, estimates a ~30% chance AI research becomes substantially automated by end of 2027 and ~60%+ by end of 2028, arguing AI doesn't need genius-level creativity to self-improve.

The latest ARC-AGI-3 scores show GPT-5.5 High at 0.43% and Claude Opus 4.7 at 0.18% — the most powerful models today remain effectively at zero on this AGI benchmark.

Google DeepMind has published a cognitive taxonomy for evaluating progress toward AGI and paired it with a Kaggle hackathon to build new benchmarks. The framework maps AI systems against human baselines across 10 cognitive abilities instead of relying on a single headline score.

Google DeepMind said on X on March 12, 2026 that a new podcast for AlphaGo’s tenth anniversary explores how methods first sharpened in games now feed into scientific discovery. The post lines up with DeepMind’s March 10 essay arguing that AlphaGo’s search, planning, and reinforcement ideas now influence work in biology, mathematics, weather, and algorithms.

Google DeepMind said on March 17, 2026 that it has published a new cognitive-science framework for evaluating progress toward AGI and launched a Kaggle hackathon to turn that framework into practical benchmarks. The proposal defines 10 cognitive abilities, recommends comparison against human baselines, and puts $200,000 behind community-built evaluations.

Google DeepMind said on X that it is launching a Kaggle hackathon with $200,000 in prizes to build new cognitive evaluations for AI. The linked Google post says the effort is part of a broader framework for measuring AGI progress across 10 cognitive abilities rather than a single benchmark.

Google DeepMind's Aletheia AI research agent solved 6 out of 10 open research-level math problems in the FirstProof Challenge as judged by expert mathematicians. The system also generated a fully autonomous research paper and solved 4 open conjectures from Bloom's Erdős database.

OpenAI CEO Sam Altman has set a new date for AGI, claiming that by the end of 2028, most of humanity's intellectual capacity could reside inside data centers rather than human minds.

DeepMind CEO Demis Hassabis proposed a concrete AGI benchmark: train an AI with a knowledge cutoff of 1911, then see if it can independently derive general relativity as Einstein did in 1915. This test targets genuine scientific discovery rather than pattern matching.