AIWhat comes after AI scaling?
An announcement for The Information Bottleneck describes a conversation with Adaptation Lab co-founder and CEO Sara Hooker about mostly static models and what it would take for AI to adapt continuously to new tasks, data and users.
TLDR
The Information Bottleneck episode announcement says the conversation with Sara Hooker explores continuously adapting AI, AutoScientist and self-improving systems. It also lists why non-verifiable tasks may be the next bottleneck and why interfaces may matter as much as models. Other topics include open versus closed models, distillation and Chinese AI labs, AI safety, cyber risk, biorisk, and what comes after Transformers and tokenization.
Combined views
1.6K
1 Source, first seen 1d ago
What comes after AI scaling?
An announcement for The Information Bottleneck describes a conversation with Adaptation Lab co-founder and CEO Sara Hooker about mostly static models and what it would take for AI to adapt continuously to new tasks, data and users.
TLDR
The Information Bottleneck episode announcement says the conversation with Sara Hooker explores continuously adapting AI, AutoScientist and self-improving systems. It also lists why non-verifiable tasks may be the next bottleneck and why interfaces may matter as much as models. Other topics include open versus closed models, distillation and Chinese AI labs, AI safety, cyber risk, biorisk, and what comes after Transformers and tokenization.