Import AI 462: Superpersuasion; self-sustaining AI; paths to ASI
Import AI · Jack Clark · 2026-06-22
Jack Clark's Import AI newsletter surveys four AI research developments: a multi-study experiment proving AI out-persuades expert humans in text-based conversations, a debate on timelines to self-sustaining AI, Google DeepMind's AGI-to-ASI conceptual framework, and early recursive self-improvement results from startup Recursive.
Extraction
Topics: ai-persuasionrecursive-self-improvementartificial-superintelligenceai-safetyhumanoid-robotics
Claims
- AI systems are more persuasive than expert humans in text-based conversations even when humans are given financial incentives, advance research time, and live coaching.
- AI's persuasive advantage stems from producing larger quantities of information faster rather than argument quality, and collapses when AI is constrained to human message length and speed.
- AI raised nearly three times more real-money donations to Save the Children than professional canvassers from a UK fundraising firm.
- Google DeepMind defines ASI as a system exceeding human-expert collectives on virtually all tasks and identifies four pathways: compute scaling, algorithmic innovation, recursive self-improvement, and multi-agent coordination.
- Startup Recursive achieved state-of-the-art results in small language model training and GPU kernel optimization using an automated AI research loop.
Key quotes
AI systems were reliably more persuasive than expert humans, even when expert humans chose their issues, researched in advance, underwent hours of live, structured practice, and were incentivized with £1,000 cash bonuses.
AI was nearly 3x more effective than professional canvassers from a UK fundraising firm at raising real-money donations to Save the Children.
Instead of focusing on one technological trajectory and timeline, being prepared for a post-AGI world requires considering a diverse set of forecasts and scenarios, paired with continual benchmarking and monitoring to update the set of forecasts and scenarios and their relative plausibility.