Industry Newscontinual learninglong term planningdeepmindrobustness
DeepMind CEO Identifies Three AGI Shortcomings
8.1
DeepMind CEO Demis Hassabis said at an AI summit in New Delhi that current AGI efforts still fall short in three areas: continual learning, long-term planning, and inconsistent performance. He noted systems are typically frozen after training, can only plan short-term, and sometimes err on simple tasks despite excelling elsewhere. Hassabis previously predicted true AGI could appear within five to ten years.
Key Points
- 1Identifies three AGI gaps: continual learning, long-term planning, and inconsistent performance on simple tasks
- 2Highlights that current models are frozen post-training and cannot adapt to new online experiences
- 3Implies researchers should prioritize continual learning, hierarchical planning, and robustness testing in models
Scoring Rationale
Senior DeepMind commentary highlights core AGI gaps, but offers limited new technical detail or actionable solutions.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
