Singularity: latest AI news

Forecasts, timelines and capability evaluations: how close is general AI?

• 1 Nov Mathematician Frank Calegari urges colleagues to adapt after OpenAI's release of over 700 AI research papers
• 1 Nov Z.ai co-founder Tang Jie calls recursive self-improvement the main research problem for reaching AGI
• 31 Oct Survey finds Chinese AI labs aim recursive self-improvement at training and serving systems, not model weights
• 30 Oct Cambricon co-founder Chen Yunji says AI-designed chips and smarter AI could form a self-improving loop
• 30 Oct Musk and Zuckerberg say 'super intelligence' for AI on earnings calls; Alphabet, Amazon, Microsoft CEOs do not
• 29 Oct Artificial Analysis launches a Progress Index putting every language model since 2022 on one capability scale
• 29 Oct DeepMind researchers model AI agent coordination, not competence alone, as the driver of abrupt economic change
• 29 Oct Goldman finds workers displaced from AI-exposed jobs take about a month longer to find new work
• 29 Oct Mento benchmark: 14 AI models research startups well, but all 13 finishers rank growth worse than a headcount rule
• 29 Oct Will Manidis and Joe Roberts publish 'The Commanding Heights', a 40,000-word essay on who captures AI's value
• 29 Oct Sam Altman tells Stripe's John Collison he is against a single superintelligence
• 27 Oct Artificial Analysis launches Intelligence Index v5, with 60% of the score from private test sets
• 18 Oct Anthropic commits 5% of its shares to a new Public Benefit Fund, pending shareholder approval
• 15 Oct Anthropic gives METR staff employee-level access and lets it publish risk assessments without veto
• 11 Oct DeepRSI benchmark gives AI agents 48 hours to improve real ML research tasks