Ochrana soukromí Zprávy 24 Leden 2026

February 5, 2026, 1:42 PM EST. An evolving chart from MIT Technology Review's METR plot continues to shape AI chatter. The graphic, first released last year, tracks near-exponential progress in certain tasks and has appeared to outpace itself with recent model releases, notably Anthropic's Claude Opus 4.5. METR cautions that its estimates carry large error bars and measure specific tasks-chiefly coding-by comparing them to human time to completion, not universal capability. Critics warn that the chart does not reflect AI increasingly general abilities, and that reading too much into one plot risks alarmism. METR's work spans risk assessment and deeper system evaluations, including studies on whether AI assistants slow developers. The group emphasizes uncertainty and the need for broader measurement as frontier models advance.