This Week in AI: Billionaires talk automating jobs away | TechCrunch
Feb 04, 2025 - techcrunch.com
The TechCrunch AI newsletter discusses recent developments in the AI industry, highlighting OpenAI CEO Sam Altman's meeting with SoftBank CEO Masayoshi Son in Tokyo. They discussed automating white-collar workflows using AI, with SoftBank planning to invest $3 billion annually in OpenAI products. The newsletter raises concerns about the potential unemployment resulting from widespread job automation and the lack of focus on worker welfare by leading AI companies. Additionally, it covers OpenAI's new AI "agent" for research, the EU's ban on risky AI systems, and a play about AI "doomer" culture.
The newsletter also mentions a new reasoning model, o3-mini, by OpenAI, and Google's X "moonshot factory" launching Heritable Agriculture to improve crop yields. A research paper from Tencent addresses "underthinking" in reasoning models, proposing a "thought-switching penalty" to improve accuracy. The newsletter introduces YuE, a music-generating model by ByteDance and Moonshot, and Anthropic's Constitutional Classifiers, a technique to defend against AI "jailbreaks," albeit with increased computational demands.
Key takeaways:
OpenAI and SoftBank are collaborating on a platform called "Cristal Intelligence" to automate white-collar workflows.
OpenAI launched a new AI "agent" for complex research and a reasoning model called o3-mini.
The EU can now ban AI systems deemed to pose "unacceptable risk" or harm, including those used for social scoring.
Anthropic developed "Constitutional Classifiers" to defend against AI "jailbreaks," though it increases computational demand.
