Details:
The reading group will meet weekly for 4 weeks, from the week of October 26th through the week of November 16th. Each meeting will be 1 hour and include free lunch or dinner, reading, and discussion. All reading will be done during the session (no outside reading required). The deadline to register is October 19th.
Topic:
When will transformative artificial intelligence arrive? This reading group examines the heated debate over AI development timelines, contrasting the increasingly common view that transformative AI or artificial general intelligence (AGI) is imminent—perhaps within 3-7 years, as suggested by AI company CEOs like Sam Altman and Dario Amodei—against more skeptical assessments that place such breakthroughs decades in the future, if at all.
We’ll examine trends in AI benchmarks while questioning their limitations as proxies for general intelligence, analyze scaling laws and compute projections to understand physical and economic constraints, and investigate the potential for automated AI research and recursive self-improvement to accelerate progress. We’ll also explore developments in reinforcement learning, trace the interplay between data availability, algorithmic breakthroughs, and compute scaling, and track how expert opinion on the topic has shifted over time.