Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
The writer is a science commentator
Science students at Harvard University already belong to an elite but, this term, there is a super-elite among them. Fifteen have been selected to ruminate on major unsolved enigmas, such as how life sprang from nonliving matter or whether it is truly possible to reverse ageing.
The Genuinely Hard Problems scheme, designed to expose bright young minds each week to the world’s biggest unanswered questions, might usefully chart a course for other institutions to follow. According to Logan McCarty, a Harvard science lecturer and dean of education who is organising the classes with the scheme’s creator, neurobiology professor Jeff Lichtman, the internet and AI have lessened the need for ambitious thinkers to acquire specialised technical skills and internalise vast quantities of information.
Instead, McCarty told me, they should be seeking to understand human society and its problems: “Prepare students to ask, ‘how can we use science’ and ‘what should we do with science’, not just ‘how to do science’.” The initiative raises profound questions about the future of science education in the age of AI, including whether there should be a greater role for the humanities and social sciences.
Conventional scientific training can often seem like a rabbit hole, with researchers burrowing deep into narrow fields. Specialist knowledge can now be digitally retrieved in seconds; AI can mine data, construct hypotheses and design experiments. On top of that, a slender scholarly lens can obscure a wider perspective. Today, some of the biggest problems facing humanity, such as climate change and energy scarcity, tend to sprawl across disciplines rather than sit snugly within academic departments.
The primary task of scientists, the Harvard educators believe, is asking the right questions, because AI can answer even difficult queries if they are well-posed; being fearless and willing to fail, with no area of science off-limits; and doing research that is meaningful and has impact, rather than chasing quick wins.
With that in mind, the GHP cohorts — 15 this semester, and another 15 in the spring — were selected from about 160 applicants purely on the strength of their curiosity, not prior achievements. Each week, a guest lecturer describes a conundrum that has defied solution: perhaps a mathematical mystery that has endured for centuries; or the uncharted link between brain structure and mental health.
Students will pick one enigma to work on during their undergraduate years and perhaps beyond. Will Harvard’s bet on those who are curious rather than always correct pay off? “We’ll know in about 10-15 years, when the first of these students wins a Nobel Prize,” McCarty says, boldly.
Nick Lane is an evolutionary biochemist and author at University College London who works on one of those Genuinely Hard Problems: the origins of life. He uses AI to probe the border between the known and unknown, unveiling new areas for research, and agrees with the central idea that science education needs to change: “In the age of AI, which sees patterns in the data we already have, making real scientific advances will put a premium on unconstrained originality, hard thinking and creativity — to see what is unknown and unthought.”
But that did not negate the importance of traditional learning, such as how to write an essay without AI, which teaches students how to structure information. That skill, he said, allows students to see where an answer might lie, and trains them to spot incoherencies and contradictions.
He also cautioned against the conceit that long-standing mysteries just need a clever person to see the light or come up with a single equation (excepting Einstein, presumably). For example, there is more to the origin of life than, say, just building a nucleotide; complex answers need experts from different disciplines to flesh out the bigger picture: “There has to be a licence to rove between disciplines, but if everyone did it we’d be in a real mess.” Without an army of scientists to keep order, of course, there would be no disciplines to cross.
Still, a 2024 Royal Society report, “Science in the age of AI”, saw that the skills and competencies of researchers needed to adapt. Its lead author, Oxford university engineer Alison Noble, said the ultimate goal was for “humans and AI [to] work together to advance our understanding of the world beyond what either could achieve alone”.
Now, that would be quite the landmark moment: AI helping humanity to finally work out how life began.