Altman invites Washington inside the AI industry

Sam Altman used a Financial Times op-ed to advocate for establishing a U.S.-led forum to set AI safety standards and regulate access to advanced models, citing Cold War-era atomic energy regulation and aviation/banking rules as precedents. This proposal emerged from the June G7 summit where Altman met with government heads. Concurrently, reports indicate OpenAI discussed giving the U.S. government a 5% stake in the company and proposed that other U.S. AI labs contribute to a dividend fund to redistribute AI wealth. The equity discussion reflects growing momentum for wealth-sharing mechanisms, though debate centers on whether benefits go directly to households or through government intermediaries.

In consumer technology, Meta's smart glasses have achieved estimated 80% market share among all smart glasses sold globally. Despite genuine advantages like GPS, live translation, hands-free recipes, and instant visual information, the devices suffer from battery life draining in hours during heavy use and limited app ecosystem (Meta apps only). A reviewer concluded they cannot yet replace phones despite Meta's aggressive positioning with Ray-Ban, Oakley, and celebrity partnerships.

Research from Thinking Machines Lab and Bridgewater Associates demonstrates that small, specialized AI models trained on expert judgment significantly outperform frontier models on domain-specific tasks. A custom Qwen model fine-tuned on Bridgewater's investment analysis expertise achieved 84.7% accuracy on their tests compared to ~50% for GPT, Claude, and Gemini variants, while costing 13.8x less. This finding challenges the assumption that frontier models always outperform specialized alternatives, suggesting companies benefit from building custom solutions for highly specialized work rather than relying on generalist models.