Jeff Dean spent 27 years at Google, rising to head of Google AI and then chief scientist before leaving the company this year, and his advice to Gen Z is not what most would expect from someone at the center of the AI race, according to Fortune. Rather than urging young people to master every new AI tool as it comes out, Dean, now cofounder and CEO of the AI focused startup DiscoveryLoop, recommends a much broader and shallower approach: skim widely instead of studying deeply.
Dean's method is to read the abstracts of many papers rather than working through a handful in full detail. He argues that skimming ten papers beats reading one thoroughly, because it gives you ten points scattered across your sense of what might be possible rather than a single deep but narrow line of thought. He has pushed the idea further, suggesting that reviewing something like 100 abstracts can surface unexpected connections between ideas that someone focused on a single paper would never notice.
That breadth, in Dean's view, is where genuinely novel solutions tend to come from, not from mastering one subfield but from spotting a link between two areas that most people never think to compare. He also advises picking problems with roughly a five year horizon, long enough to allow for a real breakthrough but short enough that the problem stays tractable rather than turning into an open ended research program.
Dean remains optimistic about what AI can do for access to expertise, suggesting it could eventually let people reach something close to PhD level knowledge in a subject without years of formal training. At the same time, he is candid that plenty of uncertainty and plenty of personal failure remain part of the process, even for someone who spent nearly three decades at the center of one of the most influential AI research organizations in the world.