Knowing without learning
They call him the “Mozart of Mathematics,” and while media claims touting him to be the “world’s smartest man” may be hyperbolic, there’s no doubt that the Australian-born mathematician Terence Tao, at 51, is at the summit not just of his profession but of scientific inquiry as a whole. A certified child prodigy, he was writing computer programs at six, and entered high school at seven. He earned his PhD from Princeton at 21 and became full professor at UCLA at 24. In 2006, he received the Fields Medal, the Nobel Prize of mathematics, for his exceptional achievements in the discipline.
As an admirer of notable scientists, I had been following Terence Tao on YouTube for some time, but one of his most recent videos caught my attention with a phrase I’d never heard before: “cognitive friction.” “Hold on, we need some cognitive friction here,” Tao was saying.
It’s the context that makes that statement significant. Let’s step back a bit to see it. Tao was talking about one of his favorite topics – mathematics and AI. An early AI adopter, Tao had enthusiastically looked to AI as a collaborator; it could perform the basic but tedious calculations while human mathematicians provide the creative direction, the kind of inspired visioning that leads to breakthroughs.
Mathematics is driven forward by solutions to longstanding problems and conjectures, such as Fermat’s Last Theorem, which stood for 350 years until it was solved by Andrew Wiles in 1994. In 2000, the Clay Mathematics Institute offered a prize of $1 million for each solution to seven of these mathematical challenges (only one, the Poincaré Conjecture, was solved in 2003 by a Russian who refused the money).
But solving these problems is more than about parlor games and bragging rights. As with Fermat, solutions open up new lines of inquiry, often going beyond the immediate horizon. The practical benefits for you and me may be seemingly few and distant, but they eventually emerge. Studies of prime numbers have led to better encryption, keeping our bank accounts and messages safe; advances in topology have applications in fields as far apart as neurology and aeronautical engineering.
One of those pesky unsolved problems has been the so-called Navier-Stokes equation, which has to do with how fluids move (in math, “fluids” mean anything that flows, including air and smoke). I could be wrong (I famously failed math at PSHS, and so became a writer), but as I understand it, Navier-Stokes appeared on the Millennium Prize list because someone had to prove that fluids would behave with a certain predictability under certain conditions – to make sure, for example, that we are simulating weather conditions accurately for the guidance and safety of planes flying in turbulence.
A few weeks ago, something earthshaking happened in MathWorld. Rival AI companies Anthropic and OpenAI claimed that they had solved the Navier-Stokes problem, along with a raft of lesser math conundrums that mathematicians had labored for centuries to unravel. Terence Tao himself had been doing some work on it, as well as a host of other aspirants who felt that they were this close to a solution, until OpenAI suddenly pulled the rug out from under them and announced itself effectively the winner. According to Scientific American, the solution runs to a staggering 166 pages, took about a hundred hours of computing time and cost as much as $40 million, including electricity. Despite all that, the proof has been challenged by mathematicians, amid charges of plagiarism, stolen data and dubious parameters.
Terence Tao said that while a problem may have been solved, it had yet to be understood. It was an answer, all right, but what exactly did people learn from it? The company rivalry in the background also turned mathematical and intellectual inquiry into a competition for bragging rights with little regard for scholarly standards and procedures. That’s when he said, in an interview, “We need some cognitive friction here.”
The meaning of “cognitive friction” has changed somewhat over the years since it was first proposed by Alan Cooper in his 1999 book The Inmates Have Taken Over the Asylum, where he saw it to be the human reaction or resistance to unpredictability. (Full disclosure: I haven’t read that book, and am using AI to supply this and other information here.) Later, product designers – the so-called UX or user-experience community – took it to mean the kludgy things that get in the way of good design, something to be removed for the smoothest operation possible. If I read it right (and again I could be wrong), Terence Tao’s first appreciation of the term ran along this line – cognitive friction was an impediment to a better and faster way of solving problems.
But now the problem was, problems were being solved by AI blazingly fast, so much so that while the end-product, the conclusion, looked like new knowledge, no human learning really took place. In solving math and other intellectual problems, the process is just as valuable as the solution.
And this is where Terence Tao’s reappreciation of his early advocacy of AI-assisted research comes in, because now Tao is saying that we need to take long pauses in our work with AI to see and understand precisely what it’s doing and to check if it’s still working the way it should, by human standards. Slow down, Tao urges, we’re moving too fast. That’s “cognitive friction” in another sense – a deliberate delay, like a checklist or a prompt, that a user has to confront before proceeding any further.
A similar term it goes by is “cognitive forcing,” which imposes necessary interruptions, especially when dealing with AI, to enable human discernment to kick in. In medicine, cognitive forcing techniques help doctors to avoid premature diagnoses. “Effortless” can also mean “mindless,” and lead to fatal errors of judgment. Cognitive forcing – check your sources, review the data, question the conclusions, explore the options, etc. – may yet be the best way to protect and retain our intellectual independence from the machine spewing out magical answers.
I’m not sure if you found this useful on a Monday morning, but it’s something we can certainly apply to our everyday life, even the trigger-happy way we respond to others on social media (for which we can’t even blame AI, but our own human bias and occasional idiocy).
* * *
Email me at [email protected] and visit my blog at www.penmanila.ph.
- Latest



















