Summary
Peter Diamandis interviews Elon Musk live at FII in Riyadh. Musk covers AI timeline predictions (any individual human capability within 1-2 years, all humans combined by ~2028-2029), energy future (rounds up to 100% solar long-term), humanoid robots (10B+ by 2040 at $20-25K each), Tesla autonomous driving (unsupervised FSD in California/Texas by mid-2025), Starship to Mars (uncrewed in ~2 years, crewed by end of decade), and Optimus robot production timeline. Musk frames AI safety as “maximally truth-seeking AI” rather than regulation.
Key Segments
- [00:01] AI improving 10x per year; individual human-level capability within 1-2 years, all-humans-combined by ~2029
- [00:03] AI existential risk: 80-90% positive outcome, 10-20% chance it goes bad (improving from earlier 80/20)
- [00:04] Population collapse as biggest long-term existential threat; South Korea could shrink to 1/3 current size
- [00:06] AI safety = maximally truth-seeking AI; criticizes “woke nihilistic philosophy” being trained into Bay Area AIs
- [00:08] Energy: currently chip-limited, will become energy-limited; less than 1% on Kardashev scale
- [00:10] Solar energy: Earth uses 1/8,000th of sun’s energy hitting its surface; long-term energy rounds to 100% solar
- [00:13] Humanoid robots: 10B+ by 2040, price $20-25K each; future of abundance with near-zero marginal cost
- [00:15] Starship to Mars: uncrewed in ~26 months, crewed 2 years after; regulatory burden is biggest impediment
- [00:18] Tesla FSD: unsupervised in CA/TX by mid-2025; 7M cars on road, 9.5M by end of 2025
- [00:20] Optimus limited production 2025, volume 2026; autonomy = $5T Tesla, Optimus = $25T Tesla
Notable Claims
- AI getting 10x better per year (revised down from 100x/year prediction in March 2024)
- Uncrewed Starship to Mars at next transfer window (~early 2027)
- Tesla unsupervised FSD to exceed human safety levels in Q2 2025
- Cybercab volume production expected 2026
- Optimus robot: “biggest product of any kind ever”
- “Universal high income” not “universal basic income” as most likely outcome
- Starship fifth mission booster capture done entirely without AI (“not bad for a bunch of monkeys”)
Guests
- Elon Musk — CEO of Tesla, SpaceX, xAI; CTO of X
Assessment
High-signal interview despite its brevity (~21 minutes). Musk’s predictions are concrete and timestamped, making them trackable: FSD mid-2025, uncrewed Mars ~2027, crewed Mars before 2030, 10B robots by 2040. The AI safety answer is notably thin — “maximally truth-seeking” is a philosophy, not an implementation plan, and the “woke AI” framing is political positioning rather than technical analysis. The energy discussion (Kardashev scale, 100% solar long-term) is standard Musk but useful as a concise statement. The population collapse segment is increasingly a Musk hobby horse that gets airtime in every appearance. Most valuable as a prediction timestamp — these specific claims can be scored against reality. The 10x/year AI improvement rate (downgraded from 100x) is a notable revision worth tracking.