Reid Hoffman stresses that widespread adoption, not hype, drives A.I.’s impact. Courtesy of Reid Hoffman
Reid Hoffman, featured on this year’s A.I. Power Index, is best known as the co-founder of LinkedIn, but in recent years he has emerged as one of A.I.’s most influential backers. Hoffman has invested in startups spanning communication, healthcare and beyond, and in January, he co-authored Superagency with tech writer Greg Beato, arguing that A.I. will enhance, rather than diminish, human agency. What excites Hoffman most is the speed with which A.I. has entered the mainstream, adopted by people outside tech for everything from relationship management to workplace productivity.
In healthcare, Hoffman is focused on lowering costs and accelerating the development of treatments for cancer and rare diseases. He helped launch Manas AI, an A.I.-powered drug discovery company that raised $24.6 million in seed funding from Hoffman, General Catalyst and Greylock Partners. Manas AI’s founding team includes Pulitzer Prize-winning cancer physician Siddhartha Mukherjee and renowned chemist Jonathan Baell. Hoffman stresses that widespread adoption, not hype, drives A.I.’s impact, while cautioning that concentration in leading models could stifle entrepreneurial dynamism and broad-based innovation.
What’s one assumption about A.I. that you think is dead wrong?
That its impact and adoption were driven by investor hype rather than widespread user adoption.
If you had to pick one moment in the last year when you thought “This changes everything” about A.I., what was it?
For all the talk about the fact that AGI, however one defines it, hasn’t arrived in a way that alters society in a way that is impossible to ignore, I’d say we’re still making progress faster than most people believed we would, in terms of model capabilities. But it was actually a cultural aspect to the story that struck me as most significant, and that involves the number of conversations I had with people who aren’t immersed in the A.I. world or even the tech world, who spoke about how substantially and impactfully they are using A.I. to improve their lives—whether that’s through using it to navigate their relationships with their spouses, or boost their productivity at work or something else that is meaningful to them.
I don’t recall a similar level of wide-ranging mainstream adoption just a couple of years into the Web 1.0 era, for example. As much as there are obvious concerns and even resistance to these new technologies, the public’s overall embrace has been eye-opening to me.
What’s something about A.I. development that keeps you up at night that most people aren’t talking about?
Like any successful technology, A.I. will be prone to lock-in effects, and this is especially true because A.I. systems are not just tools but adaptive infrastructures. On the one hand, they have all the usual path dependencies: The data used to train them, the architectures chosen and the institutional practices built around them will generally be costly to replace. But then you’ve also got the fact that A.I. systems are fundamentally learning-based systems that improve through use, which will theoretically make it even harder for challengers to unseat incumbents down the line. And obviously, this is concerning on multiple levels.
It means early design choices and deployment contexts play a major role in how A.I. evolves and whose interests it ultimately serves. And it could lead to a world where we see a decline in the kind of entrepreneurial dynamism that has traditionally been the source of new job creation, competition, innovation, and broad-based productivity growth.
