Marcus on AI ⢠11461 implied HN points ⢠23 Dec 25
- Huge bets on large language models have driven a boom in chips and data center construction, but real-world performance and trust are lagging, so those assets could become overvalued and risky.
- Multiple studies and company experiences show generative AI often fails to deliver the promised productivity gains and can sometimes harm outcomes, so itās premature to treat it as a guaranteed productivity revolution.
- Putting an entire economy or national strategy all-in on generative AI is dangerous; diversification and cautious risk management are needed to avoid big losses or calls for bailouts.