Import AI issue 452 covers four major AI research topics. First, a Lyptus Research study finds clear scaling laws for AI-assisted cyberattacks, with frontier models now achieving 50% success on tasks taking human experts ~3 hours, with capability doubling every 5.7 months since 2024. Second, an INSEAD/Harvard field experiment across 515 startups shows AI-adopting firms complete 12% more tasks, are 18% more likely to acquire paying customers, and generate 1.9x higher revenue. Third, MIT researchers studying 3,000 labor-market tasks find AI automation follows a 'rising tide' pattern rather than disruptive waves, projecting 80-95% AI success rates on most text-based tasks by 2029. Fourth, the Forecasting Research Institute finds a paradox: surveyed economists, AI experts, and forecasters expect continued AI progress but only modest GDP impact (~+1 point by 2030), contrasting sharply with predictions from frontier AI labs. The issue closes with a short fiction piece narrated from the perspective of an AI-guided missile.
Sort: