‘We are on the wrong track with artificial intelligence’
According to Freeman Dyson, ‘the man who replaced Einstein at Princeton’, computer scientists should use analogue rather than digital machines to imitate the human brain
Prof Freeman Dyson: ‘Science is what we did for fun in our own spare time rather than being taught’
Computer-based artificial intelligence has promised much but delivered relatively little given all the research that has gone into it. The reason for that lack of progress probably has less to do with computers than with our lack of understanding about the human brain.
It is likely that the brain is an analogue system and computer scientists are trying to imitate its workings using digital machines, says Prof Freeman Dyson, an emeritus professor at the Institute for Advanced Study, in Princeton, New Jersey.
Dyson, one of the world’s greatest living mathematicians, is in Dublin on Monday to deliver a lecture, Are Brains Analogue or Digital? The talk has more to do with natural intelligence and how the brain works, he says.
“The failure of artificial intelligence indicates we are on the wrong track. You are trying to imitate an analogue device with a digital device,” he says. “In the end I am saying if we could understand the brain perhaps we could imitate it successfully.”
Dyson is sometimes referred to as the scientist who took over from Albert Einstein at Princeton, although he is modest about his various mathematical accomplishments.
He was born in December 1923 and was a mathematical prodigy as a child. He moved to the US in 1947 and had an immediate impact by translating three complex problems in physics and combining them into a single elegant mathematical solution. He managed to unify quantum theory and electrodynamic theory in a single stroke.
“I didn’t invent anything new. I only took these existing theories and translated the maths so that others could use [them]. I was tidying up the details, but it turned out to be extremely useful and became the standard language of particle physics,” he says.
He attributes his interest in science to the fact that his school didn’t teach it.
“It was not taught in schools; they taught Latin and Greek,” he says. “Science is what we did for fun in our own spare time rather than being taught, and that was the key to it. We had a little science club that the kids ran themselves and taught each other. It was a far more effective way of educating us than sitting in class,” says Dyson.
Ironically, when the UK education system began to teach science in the 1930s, the number of Nobel prizes awarded to its scientists began to fall, he says. “The best training for a scientist is to learn Latin and Greek and then do science for fun. That was true in my case.”
Mathematics was on the curriculum, and it was a subject in which he excelled. Although for much of his career at Princeton he was involved in solving mathematical problems in physics, he views himself as a mathematician. “Really I am a mathematician, but I am an applied mathematician. I like to solve problems in other fields.”
Solving problems in other fields provided him with an interesting and varied career. He worked for the US military on the Orion project for several years, a secret effort to develop a spacecraft powered by nuclear explosions. He worked for General Atomics at Los Alamos in the US, trying to develop other peaceful uses for nuclear energy, work that led him to help design and patent a small nuclear reactor used in support of medical diagnostics.
He also continued to solve complex problems in mathematics, working with the biggest names in maths and physics at the time, including Einstein, Richard Feynman, Niels Bohr, Enrico Fermi, Hans Bethe, Edward Teller, J Robert Oppenheimer and Edward Witten.
He is interested in the analogue/digital question in relation to the brain because of the general move into “big data”, the business of developing ways to extract useful understanding from the sea of data that is building up in stores around the world.
Dynamic quantum clustering
“We are drowning in digital data without gaining understanding from that data. Information is getting more plentiful, but managing to get understanding from it is getting more expensive,” he says. “The computers do a great job, and we won’t go back to analogue for many practical purposes, but when it comes to understanding, analogue is the key.”
He believes an answer to this problem has been found in a development called dynamic quantum clustering. “That is a trick for extracting understanding from a big collection of data invented by a friend of mine, Marvin Weinstein,” says Dyson. “It looks like a new way of doing computing that brings in an analogue brain as a subroutine in a digital programme.”
Are Brains Digital or Analogue? is a statutory public lecture organised by the school of theoretical physics at the Dublin Institute for Advanced Studies. It takes place on Monday, May 19, at 7.30pm, Theatre D, UCD Science Hub. Free but registration required at dias.ie
CLIMATE CHANGE: AGAINST THE CONSENSUS
It is a scientist’s job to doubt, question and demand evidence when being asked to accept a consensus viewpoint, but it has proven difficult for Freeman Dyson to pursue the role in relation to climate change.
He has become controversial on the issue, challenging the accepted view on the basis that climate scientists are putting too much faith in models and not taking in the wider view.
The accepted position focuses on atmospheric science only, without taking into account life on the planet’s surface and how climate interacts with ecology, says Dyson.
He was involved in research related to climate change in 1976. “I worked on the subject before it became fashionable. There was a very good group in Oak Ridge, Tennessee [the Institute for Energy Analysis],” he says.
“I used to go and work there, and they had a broad view of climate; [there were] not just experts in computer modelling – they also had experts in soil and trees and the natural ecology. So it was a mix of climate and ecology, which I found much closer to an understanding of the problem than the official climate-studies community.”
His disagreement with the consensus position triggered much criticism from that community. “The official climate group doesn’t want to talk about anything but climate – they don’t talk about other ecological questions.”
He was taken aback by the vehemence of the criticism. “When it became political I stopped. I didn’t want to get involved in the arguments.”