ASI-LIB-010 technical frontier report

From AGI to ASI

Tim Genewein, Matija Franklin, Alexander Lerchner, Laurent Orseau, Samuel Albanie, Adam Bales, Cole Wyeth, Stephanie Chan, Iason Gabriel, Joel Z. Leibo, Allan Dafoe, Marcus Hutter, Thore Graepel, Shane Legg

This is the site anchor paper. It is not a benchmark result or a narrow systems paper; it is a map of the territory immediately after AGI. The useful move is that it treats ASI as a continuum of machine intelligence rather than as a mythic discontinuity.

Why it matters

The report identifies four pathways worth tracking:

  • Scaling AGI: continue increasing compute, data quality, context, memory, tool use, and post-training.
  • Paradigm shifts: new model classes, learning algorithms, training objectives, memory systems, or search procedures.
  • Recursive improvement: AI systems help improve the methods by which AI systems are built.
  • Multi-agent collectives: large populations of specialized agents behave like cognitive organizations with machine-speed coordination.

For a pro-ASI technical agenda, the key conclusion is practical: progress can compound through many partially independent loops. A single “AGI arrives” date is less useful than monitoring which loops are already closing.

Research agenda

Track the bottlenecks explicitly: evaluation quality, compute allocation, long-horizon autonomy, automated experiment design, agent reliability, and the rate at which AI-generated improvements transfer into the next generation of systems.