Key ideas from Terence Tao

• AI has driven the cost of idea generation in science down to almost zero, but verification and validation remain bottlenecks
• Over 50 Erdős problems have been solved with AI assistance, but success rate for any given problem is only 1-2%
• Tao predicts AI will be a trustworthy mathematical co-author by 2026, making him 5x more productive on certain tasks
• Current AI excels at breadth while humans excel at depth - future science will require redesigning research paradigms
• Mathematical progress may increasingly come from AI mapping out entire fields before humans tackle the difficult islands
• The Riemann hypothesis remains unsolved partly because we lack formal languages for mathematical strategies, not just proofs
• Hybrid human-AI collaboration will dominate mathematics longer than purely autonomous AI solutions
• Terence Tao's team developed MRI algorithms that are 10 times faster than traditional scans, now used in all modern MRI machines