Head of Amazon's AGI Lab Is Leaving the Company

When a leading AI researcher walks away from a tech giant saying AGI is "so close," it deserves attention. That is how David Luan, who headed Amazon's AGI lab in San Francisco, explained his decision to leave after less than two years. On Tuesday he posted on LinkedIn that he is leaving at the end of the week to "cook up something new."

Luan is a prominent figure in AI research. Before joining Amazon, he held leadership roles at several AI startups and was known as a specialist who could build research teams from scratch. Amazon brought him in precisely to strengthen its position in the race for frontier language models and, more broadly, for artificial general intelligence. The San Francisco lab was meant to become one of the key hubs of that effort, competing for talent with OpenAI, Google DeepMind and Anthropic, whose offices are literally next door.

But the wording Luan chose for his farewell post says a lot. He stressed that Amazon has "incredible work and opportunities to take on more areas," but immediately added that he preferred to "devote one hundred percent of his time to teaching AI systems entirely new abilities." Between the lines, one can read dissatisfaction with the pace or direction of work inside the corporation. When a researcher of this caliber says he wants to "cook up something new" outside a company with practically unlimited resources, it signals systemic problems.

And those problems do appear to exist. According to The Verge, the situation with AI products inside Amazon is far from ideal: leaks indicate the company's own employees call its internal AI developments uncompetitive. Alexa, once a revolutionary voice assistant, never made a qualitative leap in the era of large language models. The Nova family of models, which Amazon introduced as part of its Bedrock platform, has yet to generate the enthusiasm that accompanies releases from OpenAI or Google. Even Amazon's strategic investment of up to $4 billion in Anthropic looks more like an admission of its own limitations than a show of strength.

The context of Luan's exit matters also because it reflects a broader trend. Big tech corporations are increasingly losing the fight for elite AI researchers. The best specialists choose either startups, where they can move faster and avoid corporate bureaucracy, or found their own companies. Over the past year we have seen a wave of such moves: leading scientists left Google, Meta and Microsoft to launch their own projects. Amazon, already regarded as an underdog in the big AI race among the FAANG companies, can least afford such losses.

Luan's remark that AGI is "so close" deserves separate attention. Researchers have no consensus on when artificial general intelligence will arrive; estimates range from a few years to several decades. But when a practicing lab head, rather than a marketer or a venture investor, makes such a statement, it at least suggests that something encouraging is happening at the research frontier. The question is whether Luan saw enough ambition and resources inside Amazon for that leap, and judging by his decision, the answer is no.

What exactly Luan will "cook up" is not yet known. The wording hints at launching his own startup, though a move to one of the leading AI labs is not ruled out. Either way, it is a painful loss for Amazon. The company of Jeff Bezos and Andy Jassy finds itself in a paradoxical position: with one of the world's largest cloud infrastructures through AWS, enormous computing power and vast financial resources, it cannot retain the people who are supposed to turn all of that into breakthrough AI products. Money and servers, it turns out, are no substitute for a culture in which researchers feel they are moving toward AGI at maximum speed.

For the whole industry, Luan's departure is one more confirmation that the center of gravity in AI research keeps shifting from large corporations to more agile organizations. And if Amazon cannot find a way to reverse that trend, no billions invested in Anthropic will help it take a place in the front row of the new technological era.

*Meta is recognized as an extremist organization and banned in Russia.