News Summary

Miles Brundage, a former policy researcher at OpenAI, has publicly criticized the company for its approach to AI safety and alignment. In a recent document, OpenAI outlined its philosophy on developing artificial general intelligence (AGI), suggesting a continuous path of iterative deployment and learning from current AI technologies. Brundage, involved in the release of GPT-2, argues that OpenAI’s earlier caution in deploying AI models, such as GPT-2, was consistent with its current strategy and necessary at the time. He expressed concerns over OpenAI’s portrayal of past decisions and highlighted the potential dangers of downplaying caution in AI development. The discussion comes amid increasing competition and financial pressures on OpenAI, with rivals like DeepSeek emerging. OpenAI’s projected annual losses could reach $14 billion by 2026, prompting a faster product release cycle, which Brundage and others worry may compromise long-term safety.