Beyond the Resume

In late 2014, the guy who now runs a $965 billion AI company had never worked in AI.

Dario Amodei had spent a decade on physics and biology: a Princeton PhD recording how roughly 200 nerve cells in a piece of retinal tissue fire together, then a postdoc in a cancer lab. He joined Andrew Ng’s team at Baidu that November. Greg Diamos (a colleague there) said the team saw a biology guy with no machine learning background. Then Diamos read the code Dario had written at the cancer lab and saw an unusually great programmer.

At Baidu, his job was to build the best speech recognition system he could. The field’s view, in his telling, was that the algorithms needed to succeed hadn’t been invented yet. He didn’t know enough to agree, so he just tried the obvious things: make the model bigger, feed it more speech data, train it longer, and watch what happened.

Every time he made the model bigger or added more data, it transcribed speech more accurately. He figured the experts were probably right, and he was missing something, but he kept going anyway. Years later, he called finding that pattern “almost like beginner’s luck.”

He spent the next several years betting on that “luck.” His 2017 memo argued that bigger models and more data could improve AI beyond speech recognition. A 2020 OpenAI paper quantified those gains as “scaling laws”: predictable improvements from increasing model size and training data. That research helped guide GPT-3’s development. He later co-founded Anthropic, continuing to build on that approach.

I really recommend thinking about this story when making your next hire. None of that would have happened if Diamos screened for Dario the resume instead of Dario the person. He looked at what Amodei could actually do with a computer, saw raw talent, and hired based on that.

When the talent is real, the missing credentials often just don’t matter at all. And even better, those hires often don’t suffer from the same “supposed to work this way” that holds experienced teams back from seeing a breakthrough right in front of them.

Maybe your next breakthrough comes from “beginner’s luck.”