As Meta recruits AI superstars, classic pitfall awaits

Research suggests that too much talent, ego on one team can lead to conflict, 'sabotage.'

- BY JO CONSTANTZ

WITHOUT EXPERT management, teams that may be rich in talent can still fail if the chemistry is poor.

Meta Platforms Inc. is spending a fortune to assemble the brightest minds in artificial intelligence. Chief Executive Officer Mark Zuckerberg may want to note: Research suggests that packing a team with too much talent can backfire.

So far, more than a dozen engineers from OpenAI have defected to Meta, joined by notable experts from Anthropic and Google's DeepMind. Zuckerberg's wager is that by concentrating top talent and giving them unlimited resources, he can gain ground on rivals and fast-track the development of AI systems so advanced that they will approach “artificial general intelligence,” the hypothetical point at which the model surpasses human-level capacities.

As it turns out, loading up a team with multiple superstars is not always a winning strategy, as any frustrated sports fan could tell you. Without expert management, too much talent in one group can lead to diminishing returns — or even outright failure — if egos clash and the chemistry is poor enough.

“There’s this belief on Wall Street and in Silicon Valley that you just bring the most capable people, put them together and magic just happens,” said Boris Groysberg, a professor at Harvard Business School who has studied team dynamics for over two decades. “No, magic doesn’t happen. What you have in many cases is a lot of jealousy, backstabbing, sabotage.”

At Meta, keeping the superstars out of these traps is a task that falls to Alexandr Wang, the 28-year-old former chief executive of Scale AI, and former GitHub Inc. Chief Executive Nat Friedman, 48, the newly appointed leaders of Meta’s 50-person superintelligence unit.

Decades of academic studies indicate the challenges they'll face.