The Sequence Radar - Issue 948: Last Week in AI: Billions in Funding, Hundreds of Papers, One Big Leap?

This week saw new AI models from Reflection (Beam) and Mistral (Large 4), emphasizing efficiency and European sovereignty respectively. OpenAI released numerous AI-generated math manuscripts, sparking debate about verification and the acceleration of scientific research. Significant funding rounds for TypeSafe ($870M) and Arena ($200M) highlight the growing investment in specialized AI and its evaluation.

• Reflection introduced Beam, a 501 billion parameter mixture-of-experts model focused on reinforcement learning and inference efficiency.
• Mistral released Large 4 (Le Chonk), a trillion-parameter multimodal model emphasizing European infrastructure and sovereignty.
• OpenAI published over 700 AI-generated math manuscripts, raising questions about verification and the acceleration of research.
• TypeSafe secured $870 million in Series A funding for its Jev model, which produces typed decisions with confidence estimates.
• Arena raised $200 million in Series B funding and introduced an Alignment Index to evaluate model behaviors beyond task completion.
• The AI industry is seen as separating intelligence into distinct economic problems: capability production, efficient delivery, control retention, and behavior verification.
  https://bender.layer3.press/articles/68f16845-3654-4177-abaf-a71faead5af1