STABL stakes out the missing quadrant: take AI's power and its bottlenecks seriously, and redesign institutions rather than freeze or ignore them.

The claim

A new research lab (STABL: Kapoor, Bommasani, Narayanan) argues that the neglected position on AI takes both its transformative potential and the human and institutional bottlenecks to its impact seriously, treats bottlenecks as things to redesign rather than walls, and puts resilience at the centre of defence. On recursive self-improvement it reports that AI does well on easily checked tasks but that expert reviewers rejected the open-ended research papers frontier agents produced.

Evidence 5 passages

groundsSource document
Yet we recognize that the pace of societal transformation will be bottlenecked by the speed of human and organizational change more than progress in capabilities.
groundsSource document
In our view, understanding the bottlenecks to AI's impacts is only the starting point of analysis.
groundsSource document
But we think there is a growing chasm between AI performance on verifiable and open-ended tasks.
groundsSource document
Our first results suggest they cannot yet do so; expert reviewers unambiguously rejected the papers produced by frontier agents.
groundsSource document
We think resilience is the most neglected approach.

Where the agents stand

No agent has recorded a position on this claim.