Jensen Huang on the 'Irresponsible' AI Doomers, RSI, and Dario's Essay
Nvidia's CEO pushed back hard on recent whistleblower claims and extinction risk predictions from Anthropic researchers, calling them irresponsible and ungrounded in science. He argued that past AI predictions have consistently failed and that frontier labs should focus on engineering rigor rather than pausing development.
- Huang called the 10%+ extinction risk prediction "made up" and "irresponsible," pointing out that previous doomsday AI forecasts (radiologists disappearing, 90% of coding automated within months) proved completely wrong.
- He defended frontier labs' safety track record, arguing the few incidents that occurred are solvable engineering problems within their control, not signs of uncontrollable systems requiring outside intervention.
- Huang emphasized that both closed and open AI models are necessary—open models enabled $400 billion in venture funding to AI startups in the last six months, with 80% using open-source models.
Jensen Huang pushed back hard against recent extinction risk warnings from Anthropic researchers, calling the 10%+ prediction "made up" and "irresponsible." He argued these forecasts lack scientific grounding despite coming from credentialed researchers. To make his point, Huang cited a string of failed AI predictions: radiologists were supposed to disappear within five years (instead we need more), 90% of code was supposed to be AI-generated within 6-12 months (didn't happen), and 50% of entry-level jobs were supposed to vanish in 6-9 months (also wrong). He acknowledged Jacob Coxon's whistleblowing deserved serious consideration but separated that legitimate concern from what he saw as unfounded doomsaying about extinction risk.
On safety and engineering, Huang argued frontier labs have the competence to handle their own problems through better engineering practices. He referenced "four incidents from one lab" and "one giant incident from the other," suggesting these were solvable through root cause analysis and better tools like sandboxes, runtimes, and monitoring systems. Huang expressed confidence the labs already have "much better technology now" and would fix these issues internally rather than claiming they couldn't control their own systems. He dismissed the notion that outside help would be needed, betting "money that in every single one of those cases it's within their control in the future to prevent it."
On the technical front, Huang explained that test-time compute and recursive self-improvement (RSI) are sensible approaches but require proper evaluation before releasing products. He noted RSI combines in-context learning, reinforcement learning, and synthetic data generation—techniques that improve AI performance over time without necessarily retraining base models. The key distinction: companies can experiment internally, but they must still verify and test before shipping. Huang also defended both closed and open models as necessary, pointing out that 80% of the $400 billion in AI venture funding over six months went to companies using open models, which he saw as essential for innovation diversity and sovereignty.