What’s one underrated big idea?
Tacit knowledge — the idea that, as the Hungarian-British polymath Michael Polanyi put it, “We know more than we can tell.” Basically, it means that any type of complex ability — speaking a language, painting a picture, riding a bike, sailing a boat or simply walking — depends on skills that we cannot, or at least find it difficult to, articulate. We do not acquire these skills by absorbing information or learning rules but instead through experience.
The concept of tacit knowledge helps us understand our own abilities — not just what we can do, but how we come to acquire and develop our own capabilities. This, in turn, can help us identify what is of irreplaceable value in the process of learning and using human skills — not what we can do that AI cannot, but also what humans ought not offload to machines. Something that worries me about AI is that, regardless of whether it can perform a skill at or beyond the level of a competent human, it encourages us to think about our existing capabilities as if they were mere algorithms for getting things we want. Tacit knowledge reminds us that there’s something much more to the capacities we cultivate throughout our lives.
What’s a technology you think is overhyped?
I think a lot of the promises we hear about the outcomes of particular applications of AI are overhyped. AI is remarkable — it deserves a lot of hype, maybe in some instances more hype. But in a lot of the marketing and press, and unfortunately even some more scholarly discussion, you hear a lot of silly things like “AI will solve medicine” or “cure cancer” or render radiologists or even doctors completely superfluous.
The trouble comes from looking at cases where AI has solved real problems — predicting protein structures, disproving long-standing mathematical conjectures — and generalizing them to situations in which what is at issue aren’t problems in anything like that mathematical sense. Medicine isn’t a “problem” to be solved — it’s a domain of expertise that depends on tacit knowledge and, of course, technology, to address a whole variety of problems — many of which aren’t really even “problems” so much as conditions, afflictions that come from being the kind of mortal creatures we are.
What could the government be doing regarding technology that it isn’t?
Assuming intellectual and moral leadership in what is turning out to be among the most important debates of our time: What does AI mean for our future? The remarkable response to the recent papal encyclical shows how much appetite there is for this.
We can point to historical moments where similar things actually happened. One was during the 1960s and 1970s, when there were a lot of concerns about environmental degradation, the automation of work, nuclear risks. In response to such concerns — as well as its own sense of impotence faced with a distended executive branch and an expansionary presidency — Congress implemented a number of reforms to help it reassert a role in technically and socially complex areas of public policy.
Another was the President’s Council on Bioethics, established by George W. Bush in the early 2000s and first led by Leon Kass, in response to the ethical concerns posed by rapid advances in biotechnology — especially embryonic stem cell research and the potential for human cloning. Its purpose was to study the ethical dimensions of cutting-edge biotechnology and provide policy options and recommendations to inform the president’s decision-making. This created a deliberative forum at the national level for grappling with difficult questions that were at once technological and moral, scientific and political and which had enormous practical consequences for everyone.
What book most shaped your conception of the future?
Daniel Bell’s The Coming of Post-Industrial Society. There’s a lot in the book that feels dated, and some of it is flat-out wrong. I also don’t share all of Bell’s underlying philosophical premises. But it is a tremendous work that not only capture so much of the thinking of a time that has a lot of parallels with our own — it came out in 1973 — but also provides a framework for thinking about the nature and consequences of the shift from a society in which manual labor was central — not only economically but also socially, culturally and politically — to one in which scientific knowledge itself had become an essential proactive force.
I think we’re still reeling from the consequences of the changes Bell was already seeing over half a century ago — and which are likely to accelerate in the age of AI. Bell did not, as many of his critics claimed, imagine that a transition to a “post-industrial society” would eliminate manual labor or its importance — any more than industrialization eliminated agriculture — nor that the transition would be seamless or easy, much less usher in a liberal utopia. On the contrary, his discussion of both the dangers of technocracy and the fractious politics of meritocracy, higher education and the welfare state offered a remarkably prescient warning to his contemporaries.
What has surprised you the most this year?
AI’s apparent ability to solve some real, hard math problems — like refuting mathematical conjectures, including Erdős problems. There have been many claims and expectations about how AI will be able to enable, or even itself advance frontier research — rather than, say, just making calculations more efficient — in recent years. But a lot of that has been hypothetical. With these results, as with AlphaFold a few years ago, we have concrete examples that show how AI might really be able to do that going forward, at least in some domains.
At the same time, these examples illustrate my earlier point. Highly-skilled mathematicians were needed not only to check and validate these results, but also to interpret and refine — really even just to make sense of — what the AI was doing. I think this supports the “force multiplier” idea, that instead of replacing people, AI will be a very powerful complementary tool at least for those with deep experience and expertise — tacit knowledge.