It wasn’t so long ago that artificial intelligence first beat world champions in chess, then masters of GO and later Jeopardy. Those highly published stunts certainly got the attention of the public.
What wasn’t so obvious at the time was that early versions of AI were being developed for the defence industry bent on developing powerful computer systems for identifying, analyzing and devising countermeasures to potential threats such as incoming missiles.
In 2022, ChatGPT introduced an AI system that harnessed large-language models to provide natural spoken or written language links between users and computers, and hence easy access to information and analyses enabled by huge databases.
Almost overnight, users began to use AI as their preferred site for looking up information, writing essays, creating grant proposals and, lately, designing, supervising and analyzing research data — even creating music, visual art and poetry.
On the more practical side, AI is commonly used to analyze imaging studies such as X-rays, CT scans and MIRs and more recently has proven as capable as most medicine specialists in making diagnoses, prompting and analyzing laboratory studies and even suggesting treatment options.
One article from Feb. 9 written by Gina Kolata for the New York Times was provocatively titled, “AI Is Making Doctors Answer a Question: What Are They Really Good For?” — A good question for even the most seasoned and experienced physicians now and more so in the near future.
Just a few years ago, most physicians saw AI, at best, as an aid, then a partner, and now many feel threatened by AI’s growing clinical skills at eliciting and making sense of the history and laboratory findings — but as yet unable to carry out a physical examination — so far.
And if that’s what’s happening to physicians, the same is surely instore for lawyers, financial consultants and even mathematicians. For example, many promising students looking at a career in mathematics are put off these days because of AI’s mastery of all but the most challenging mathematics.
A month ago, the New York Times highlighted the latter issue in an article written by Siobhan Roberts titled, “These Mathematicians Are Putting AI to the Test.”
A group of four highly regarded, award-winning mathematicians set about devising very challenging mathematical questions for AI to solve.
The result? AI failed each challenge.
Even so, it had taken four brilliant mathematicians to come up with challenges tough enough for AI to fail.
However, my guess is that had AI been trained on similar challenging math problems, it wouldn’t have taken long before AI mastered other questions posed by top-notch mathematicians; and with the advent of quantum computing linked to AI, the combination will best even Fields Medal winners, one of the top prizes in mathematics.
But as with the challenges from mathematicians, so too will AI eventually win out in response to questions posed by human experts, whoever they might be, provided it is trained on similar, but not the same, questions.
There’s a related issue here. The more physicians come to depend on AI, the more their clinical skills may wither.
Learning to assess clinical cases can be very challenging; good examples are the weekly clinical pathological conferences cases discussed in the New England Journal of Medicine, in which physicians unfamiliar with the cases are challenged by complicated, sometimes esoteric cases.
Most of the time, invited guest discussants get the answer right, but not always. So far, AI does remarkably well on similar case material.
The future problem for physicians learning critical analytical skills will be how to acquire those skills in the age of AI.
That’s the thing with AI: it learns from its mistakes. The more AI is exposed to different challenges, such as cases in medicine and dentistry, the better it gets. Properly trained, it learns from far more cases than any one physician or dentist will in a lifetime.
Looking back at physics in that Camelot period from 1900 to 1930, a period that ushered in quantum physics and general relativity, it’s hard for me to see how AI would have come up with Einstein’s general relativity — it was such a mind-blowing creation of his imagination. Perhaps the same might be said for Schrodinger’s and Heisenberg’s solutions for the subatomic world.
Paul Dirac, one of the most brilliant theoretical physicists and mathematicians of the 20th century by wide agreement among his colleagues, once spoke of Einstein’s general relativity as a creation so novel that no one else would have come up with for 100 years or more, whereas many advances, and it might be added Nobel prizes here, are for work, that given a few more years, others would have come to the same conclusion and result — not Einstein’s masterpiece.
Einstein was one of a kind, according to Dirac, like some of the greatest artists, including some of whom created the magnificent cave art in Europe in the period between 10-30,000 years ago.
Dr. William Brown is a professor of neurology at McMaster University and co-founder of the InfoHealth series at the Niagara-on-the-Lake Public Library.
