Will machines ever be intelligent?
TL;DR
Microsoft researchers Subutai Ahmad and Nicolò Fusi join Doug Burger to debate whether today's AI systems are on a path toward genuine intelligence.
Key Points
- The conversation centers on comparing transformer architectures with the human brain, especially around continual learning and energy efficiency.
- Current large language models consume orders of magnitude more energy than biological neural networks and suffer from catastrophic forgetting when retrained.
- The researchers see no clear roadmap from existing transformer models to general human-level intelligence.
Nauti's Take
A Microsoft Research podcast that critically questions its own core technology – that is either brave science communication or very clever expectation management ahead of the next product announcement. Probably both.
The gap between 'impressive in conversation' and 'actually intelligent' remains vast, and nobody has convincingly explained how scaling alone is supposed to close it. Anyone announcing AGI timelines of 2–5 years should listen to this episode twice.
Sources