Storia in 2 fonti

AI agents struggle to perform original scientific research

Among the many predictions about the future of artificial intelligence is that models will one day be able to conduct scientific research on their own, leaving humans out of the equation. Already, they can write code, run experiments and search scientific literature, but carrying out open-ended research would require a significant leap in ability.

Raccontata da

nature.com

techxplore.com

Confronto fonti

2 prospettive sulla stessa storia

AI · summaries

techxplore.comStai leggendo21 h fa

AI agents struggle to perform original scientific research

Frontier AI agents failed six-day research tasks with rejection scores 2/6 and 1/6; flawed experiments, weak reasoning, and poor time management exposed critical capability gaps. For tech leaders: autonomous scientific R&D isn't yet feasible—position AI as assistants for routine work, not independent research.

originale

nature.com2 g fa

AI isn’t ready to research itself

Sakana AI's autonomous research system scored 1-2/6 on novel tasks, excelling at experiments but weak on hypothesis refinement. Current AI amplifies research execution but breakthrough discovery still requires human problem-solving.

Leggi questa versione →

Timeline cronologica

giovedì 13 agosto 2026·nature.com
AI isn’t ready to research itself
An agentic system successfully developed concepts from two computer-science papers — but the original authors were not impressed.
venerdì 14 agosto 2026·techxplore.com
AI agents struggle to perform original scientific research
Among the many predictions about the future of artificial intelligence is that models will one day be able to conduct scientific research on their own, leaving humans out of the…