Dario Amodei, CEO and co-founder of Anthropic, has disclosed the two core convictions that motivated him to leave OpenAI and create one of its most formidable competitors: scaling laws and safety. In a candid discussion on investor Nikhil Kamath's podcast WTF Is, Amodei traced his departure back to 2019, when early experiments with GPT-2 began revealing insights that many of his colleagues were not prepared to accept.
The Discovery of Scaling Laws
"You find incredible increases in performance," Amodei said, describing the effects of scaling models with more data and computational power. "There were a lot of folks inside and outside who didn't believe it at all." Amodei joined OpenAI just a few months after its founding in 2015 and eventually led all research there. However, as GPT-2 and GPT-3 took shape, he and a small group of colleagues developed a distinct vision that OpenAI's leadership did not fully endorse.
Scaling Laws: A Minority View in 2019
Amodei's first conviction was deceptively simple: larger models with more data and compute lead to steep performance improvements. What he calls the "scaling laws" are now widely accepted in AI, but in 2019, they were a minority perspective. He and his co-founders strongly advocated for this approach within OpenAI. Leadership eventually came around, but not entirely on their terms.
The Safety Concern
The second conviction was weightier. If AI systems were to evolve into general cognitive tools capable of reasoning and problem-solving at the level of the human brain, then the manner of their construction was critically important. "The economic implications are gonna be enormous. The geopolitical implications are gonna be enormous," Amodei stated. "It's gonna transform how the world works."
He did not believe OpenAI had a genuine commitment to AI safety. Despite the company's rhetoric about responsible AI development, Amodei was unconvinced of its deep dedication to getting it right.
Taking Action
Rather than remain and argue, Amodei drew his own line. "Don't argue with someone else's vision," he said. "If you have a strong vision and you share that vision with a few other people, you should just go off and do your own thing." In 2021, Amodei co-founded Anthropic with his sister Daniela and several former OpenAI colleagues. The company has since raised billions from Amazon and Google, and its Claude models now compete directly with ChatGPT across enterprise and consumer use cases.
About the Author: The TOI Tech Desk is a dedicated team of journalists committed to delivering the latest and most relevant news from the world of technology to readers of The Times of India. Their coverage spans gadget launches, reviews, trends, in-depth analysis, exclusive reports, and breaking stories impacting technology and the digital universe.
