- AI is nearing human-level cognitive ability faster than many realise, says Anthropic CEO
- Dario Amodei warns society is unprepared for imminent AI risks and disruptions
- Amodei shifted from biology to AI, seeing potential to tackle complex problems
Artificial intelligence is approaching human-level capability far faster than most people realise, according to Anthropic CEO Dario Amodei - and society is dangerously underprepared for what comes next. Speaking on Nikhil Kamath's podcast "WTF Is", Amodei described the current moment as standing on the shore while a massive wave gathers in the distance.
"It's as if this tsunami is coming at us," he said. "It's so close, we can see it on the horizon, and yet people are coming up with explanations like, 'Oh, it's not actually a tsunami, it's just a trick of the light.'" In his view, AI systems like Anthropic's Claude are nearing the level of human intelligence in cognitive tasks, yet there remains little broad public recognition of the risks and disruptions that may follow.
Biology to Building Intelligence
Amodei did not begin his career in Silicon Valley. Trained as a physicist and biophysicist, he initially set out to cure disease. Studying biology, however, left him overwhelmed by its complexity. Proteins splice differently, modify themselves, and interact in intricate combinations. "I started to despair that it was too complicated for humans to understand," he said.
The breakthrough moment came nearly 15 years ago, when early neural networks such as AlexNet began showing promise. AI, he realized, could potentially scale far beyond individual human cognition and tackle the complexity of biology itself. He went on to work with Andrew Ng at Baidu, later at Google, and then joined OpenAI shortly after its founding. Eventually, differences in vision led him and colleagues to launch Anthropic, with a sharper focus on safety and alignment.
The Power of Scaling
At the heart of modern AI progress are what researchers call "scaling laws." Amodei simplifies them with a chemical analogy. "If one ingredient is missing, the reaction stops," he said. "For AI, those ingredients are data, compute, and model size. If you increase them together, what you get out is intelligence."
Five years ago, models could not write essays, implement software features, generate videos, or analyse footage and answer detailed questions. Today, they can. And the improvement curve has not flattened.
Governance and Risk
Despite leading one of the world's most advanced AI labs, Amodei says he is "somewhat uncomfortable with the concentration of power in AI."
Anthropic created a Long-Term Benefit Trust to appoint most of its board, composed of financially disinterested individuals. The company also advocates for AI regulation, even when it may limit its own commercial upside. "Warning about AI risks is not a good marketing strategy," he said. "But we do it because we think it's necessary."
Essential Business Intelligence, Continuous LIVE TV, Sharp Market Insights, Practical Personal Finance Advice and Latest Stories — On NDTV Profit.