A brilliant neuroscientist who may have the key to creating true artificial intelligence.

Original author: Shaun Raviv
  • Transfer
The principle of the free energy of Karl Friston can be the most comprehensive idea since the theory of natural selection of Charles Darwin. But to understand it, you need to look into the mind of Freestone himself.

When King George III of England began to show signs of acute manic behavior towards the end of the board, rumors of a king’s insanity quickly spread among the people. One legend says that George tried to shake hands with a tree, considering that he sees the Prussian king before him. The other described how he was secretly transported to a building in Queen Square in London’s Bloomsbury district to treat him there with his subjects. It is also alleged that his wife, Queen Charlotte of Mecklenburg-Strelitz, rented a whole cellar in a local pub to store supplies for the king while he was under the care of doctors.

After more than two centuries, this story about Queen Square is still often found in London guidebooks. True or not, this place over the years has been tailored to it. A metal statue of Charlotte stands on the northern edge of the square; the pub on the corner is called "The Royal Storeroom"; and a quiet public garden on the square is surrounded by people who work with the brain and people who need to work on the brain. The National Hospital of Neurology and Neurosurgery - where modern royal persons can also go for treatment - Queen Square rises in one of the corners, The neurobiology research laboratories of the University College in London designate its perimeter. Last July, dozens of patients of the neurological department with their relatives spent a whole week of perfect weather in a quiet park on wooden benches on the edge of the lawn.

On a typical Monday, Karl Freeston arrives in Queen Square at 12:25, and smokes a cigarette in the garden next to the statue of Queen Charlotte. Freestone, who has a slightly stooped figure with thick gray hair, is the scientific director of the legendary Functional Imaging Laboratory at University College London, and is known to all FIL employees. Finished with a cigarette, Freestone passes to the west side of the square, enters the building of limestone and brick, and goes to the auditorium on the fourth floor, which can be from two to two dozen people looking at the white wall waiting for him. Freestone likes to come five minutes later so that everyone else is already assembled.

His welcome words for this group of people may well be the first significant statement of the day, since Freestone prefers not to talk to other people before noon. (Therefore, at home he communicates with his wife and three sons with the help of conditional smiles and chuckles). He also rarely meets with one-on-one people. He prefers open meetings like this, where students, postdocs and members of the public who want to get Friston's expert assessment - and there are ridiculously many such people in recent years - have the opportunity to turn to his knowledge. “He believes that if a person has an idea or a question or a project, the best way to find out about it is to gather the entire group and listen to it, so that everyone has a chance to ask a question and participate in the discussion. Then what one person finds out everyone else will find out, ”says David Benrimo, a psychiatrist and intern at McGill University, who was a student of Friston for a year. "It `s very unusual. Which is typical for Karl. "

At the beginning of each meeting on Monday, everyone in a circle asks their questions. Freestone walks slowly in circles, listens to people — his glasses slide to the tip of his nose, because he always tilts his head to see the speaker. Then he responds to the questions asked by turns for several hours. “The gentleman of the Victorian era , with Victorian manners and tastes,” as one friend of Freestone described, he answers even the most stupid questions politely and with a quick reformulation. Such question-and-answer sessions — which I started calling “ask Karl” meetings — are remarkable examples of endurance, memory, a broad outlook, and creative thinking. Often they end when Freestone leaves for his tiny metal balcony, protruding from his office, for another smoke break.

For the first time, Freestone became a legend in academia, having developed many of the most important tools that allowed science to study the brain. In 1990, he invented statistical parametric markup, a computational technology that allows, as one neuroscientist said, “shove” brain images into one standard form, so that researchers can make comparisons of the activity occurring inside different skulls. From this technology, voxel morphometry has grown, an imaging technology used in one famous study to demonstrate that the back of the hippocampus of London taxi drivers grew as they gained knowledge (in order to get a taxi driver license in London, drivers need to learn 320 routes and many attractions within 6 miles from Charing Cross.

The study, published in the journal Science in 2011, used another technology to analyze brain snapshots created by Freestone — dynamic causal modeling — to determine if people with severe brain damage have minimal consciousness activity, or only vegetative consciousness.

When Freestone was accepted as a member of the Royal Scientific Society in 2006, he was described in the scientific world as a “revolutionary” influence on the study of the brain, mentioning that more than 90% of published works related to brain imagery used the methods he had invented. Two years ago, the Allen Institute for the Study of Artificial Intelligence, which manages the pioneer of AI Oren Etzioni, calculated that Friston is the most quoted neuroscientist in the world. His Hirsch Index- the metric used to measure the influence of the publications of the researcher is almost twice as high as that of Albert Einstein. Last year, Clarivate Analytics, which for two decades had successfully predicted future Nobel Prize winners in science, placed Freestone among the three most likely winners in the categories of physiology or medicine.

It is noteworthy that a small part of the researchers who make the pilgrimage in order to meet with Freestone, want to talk with him about the images of the brain. For the ten days of this summer, Freeston gave advice to astrophysicists, several philosophers, programmers working to create a more personalized competitor Amazon Echo, head of AI department of the largest insurance company, neurobiologist developing improved hearing aids and a psychiatrist, whose startup uses machine learning to help treat depression . And most of them came to understand something completely different.

For the past ten years, Freeston has devoted most of his time and energy to the development of an idea, which he calls the "principle of free energy." (Freeston describes his work related to snapshots of the brain, just as a jazz musician would describe his work as a librarian). Freeston believes that this his idea describes no more, no less as the principle of the organization of all life, including the intellect. “If you are alive, what behavior should you demonstrate?” - he is trying to answer this question.

The bad news: the principle of free energy is incredibly difficult to understand. So hard that whole rooms of very, very smart people tried to do it, and could not. There is even a Twitter accountwith 5,000 subscribers, who only does what makes fun of his vagueness, and almost all the people with whom I discussed it, including researchers whose work depends on him, told me that they did not fully understand him.

However, often the same people hastily add that the principle of free energy, in its essence, tells a simple story and solves a simple riddle. The second law of thermodynamics says that the Universe tends to increase entropy, to decay, however, living beings violently resist it. Every morning we wake up, practically the same person we were the day before, with a clear separation of cells and organs, between us and the rest of the world. How does this happen? Freestone’s free energy principle says that all life on all scales of organization — from individual cells to the human brain with billions of neurons — drives a universal imperative that can be reduced to a mathematical function. He says that to be alive is to act in such a way to reduce the gap between your expectations and the information coming from your senses. Or, in the words of Freestone, to minimize free energy.

To imagine the potential implications of this theory, you only need to look at the people who appear on the threshold of FIL on Monday morning. Some of them want to use the principle of free energy to unite theories of reason, provide a new basis for biology and explain life. Others hope that this principle will finally provide psychiatry with a functional understanding of the brain. Others come with the desire to use the ideas of Freestone to break the deadlocks of AI research. But they all have one common reason for being here, the fact that the only person who truly understands the principle of the free energy of Karl Freeston is Karl Freeston.

In the freestone office

Freeston is not only one of the most influential scientists in his field; he is also the most prolific. At 59, he works every evening and every day off, and has published more than 1,000 scientific papers since 2000. Only in 2017, he was the lead author or co-author of 85 publications - that is, about one every four days.

But from his point of view, such an exhaust is not only the result of an ambitious work ethic, but also a sign of a desire for hard escapism.

Freeston draws a carefully guarded line between the outside world and the inside, protecting the latter from invasions, many of which, it seems, are connected with "concern for other people." He prefers being on stage to private conversations, keeping other people at a comfortable distance. He has no cell phone. He always wears blue shirts, which he buys in the store sales for two pieces. He considers violations of his weekly routine in Queen Square to be “unnerving,” so he tries to avoid meeting other people at, say, international conferences. He does not like to defend his ideas.

At the same time, Freeston clearly and frankly talks about why he is engaged in science. He finds it incredibly soothing - something similar to smoke breaks - if he manages to immerse himself in a difficult task, which takes weeks to solve. He expressively writes about his obsession with finding ways to integrate, unify and simplify world noise, to which he had a tendency as a child.

Freeston believes that his path to the discovery of the principle of free energy began on a hot summer day when he was 8 years old. He and his family lived in the English walled city of Chester near Liverpool, and somehow his mother sent him to play in the garden. He turned the old log over and found several woodlice beneath it — small bugs with an external skeleton resembling an armadillo — which randomly moved, as it seemed to him at first, in search of shelter and darkness. After watching them for half an hour, he concluded that they were not really looking for a shadow. “It was an illusion,” says Freeston. - Fantasy, which I decided to consider.

He realized that the movements of wood lice did not have any specific purpose, at least not in the sense in which a person has a goal, because of which he gets into the car and drives on business. The movements of the creatures were random; they just moved faster, warmed up by the sun.

Freeston calls this his first scientific conjecture, the moment when "all these unnatural, anthropomorphic explanations of purpose, survival, and all that, simply faded into the background," he says. - And I just had to watch what was happening. In a sense, it could not have happened otherwise. ”

Freestone's father was a civil engineer who worked on building bridges all over England, and his family moved after him. In the first ten years alone, Friston attended six different schools. His teachers often did not know what to do with him, and he won most of his fragile self-respect by solving problems himself. At 10, he developed a self-corrected robot that, in theory, could move on an uneven surface, transferring a glass of water and using self-adaptive feedback actuators and mercury levels. The school even invited a psychologist to find out from the boy how he came to such an idea. "You are very clever, Karl," Freiston assured his mother, and not for the last time. “Don't let anyone say otherwise.” He said he did not believe her.

As a teenager, Freestone experienced another moment, similar to the observation of woodlice. He returned to his bedroom, finished watching TV, and noticed a blooming cherry outside the window. He was suddenly visited by a thought that has not let go of him since. There must be a way to understand everything, starting with nothing, he thought. “If I can start from one point in the whole Universe, can I bring out everything else that I need?” He lay on the bed for hours, making the first attempts in this direction. “Obviously, I completely failed them then,” he says.

By the end of high school, Freeston and his classmates became subjects in an early experiment related to the psychological investigation of personality using computers. They were asked questions, the answers to which were punched in punched cards, and conducted them through computers to find the perfect career for schoolchildren. Freeston described how he liked the development of electronics and loneliness in nature, so the computer offered him the job of a television antenna installer. This option did not seem right to him, so he went to the school career counselor and said that he would like to study the brain in the context of mathematics and physics. The consultant told Freestone that he should become a psychiatrist, which, to Freestone’s dismay, meant that he would have to study medicine.

Freestone and the consultant confused psychiatry with psychology, which he probably should have done in reality. But the mistake turned out to be useful, since she directed Freestone on the path to the study of mind and body, towards one of the experiences that had most strongly shaped his life.

After graduating from medical training, Freiston moved to Oxford and spent two years as an intern at Littlemore, a Victorian hospital. The hospital was founded as part of the “lunacy act” of 1845 [after which people suffering from mental illness were treated more as patients in hospitals and less as criminals and outcasts huddled in “shelters” / approx. perev.], and was originally intended to help transfer all the "poor lunatics" from working houses to hospitals. By the mid-1980s, when Freeston got there, it was one of the last shelters left in the backyards of English cities.

Freestone was assigned a group of 32 chronic schizophrenics, the most unpleasant patients in Littlemore, whose treatment for the most part simply meant a conclusion. For Freestone, who recalls his patients with obvious nostalgia, this became a visual aid for how easily the connections of the brain break. “It was a great place to work,” he says. “A small community of intense and colorful psychopathology.”

Twice a week, he conducted 90-minute group therapy sessions, in which patients understood their diseases together - something that resembles today's meetings “ask Karl”. The group included colorful characters who, even 30 years later, still inspire Friston's thinking processes. There was Hilary (the names of the patients were changed) who looked like she could play the senior chef on the show. "Downton Abbey ", however, before getting into Littlemore, who decapitated his neighbor with a kitchen knife, because he was under the impression that he had turned into an evil crow like a man.

There was Ernest, who had a weakness for Pastel cardigans from Marks & Spencer and sneakers to them in tone, and the former "the most incorrigible and unbridled pedophile of all you can imagine," says Friston.

Robert was also there, an agile young man who could study at the institute if he had not suffered from serious schizophrenia. He was obsessed with the crap of angels; he argued whether this substance was a curse or a blessing, whether it was visible to the human eye, and that similar questions did not occur to other people. Freeston believed that the very concept of angelic shit was a miracle. She showed the possibilities of people with schizophrenia to create concepts that are not so easily accessible to people with a brain that functions in a more usual way. “It's extremely difficult to come up with something like angels shit,” says Friston with something like admiration. "I would not have succeeded."

After Littlemore, Freestone spent most of the early 1990s, using a relatively new then technology - positron emission tomography - to try to understand what was happening in the brain of people with schizophrenia. Along the way, he invented statistical parametric markup. Freeston was firmly convinced that technology should be freely shared, and not patented and monetized (which was unusual for that time), which explains why this technique has spread so widely. Freeston could fly to the other end of the world — to the National Institutes of Health in Bethesda, for example — to share technology with other researchers. “I literally took a plane with a biometric tape, took it there, downloaded it, spent the whole day trying to make it work, taught someone to work with it, and then he went home to rest, ”says Friston. “This is how open source software worked at the time.”

In Queen Square Freeston came in 1994, and for several years in a row its office in FIL was located just a few meters from the Gatsby Computational Neurobiological Department. This department — in which researchers study theories of perception and learning from living things and machines — was then led by its founder, cognitive psychologist and computer scientist Joffrey Hinton. FIL gained a reputation as the main laboratory in the field of brain snapshots, and Gatsby became a training camp for neuroscientists interested in using mathematical models to study the nervous system.

Freestone, like many others, became infected with Hinton's “children's enthusiasm” for completely non-child statistical models, and became friends with him.

Over time, Hinton convinced Froiston that the best way to imagine the brain was to consider it a Bayesian probabilistic machine. This idea originates from the XIX century and the works of Hermann von Helmholtz , and consists in the fact that the brain organizes calculations and sensations using the probabilistic method, constantly issuing predictions and adjusting the point of view based on input data from the senses. In accordance with the most popular current view, the brain is a “pin machine”, seeking to minimize “prediction errors”.

In 2001, Hinton left London, settling at the University of Toronto, where he became one of the most important figures in the field of artificial intelligence, laying the foundation for most modern research in the field of in-depth training.

But before his departure, Freestone last came to Gatsby to visit a friend. Hinton described a new technology invented by him that allows computer programs to more effectively emulate human decision-making — it was the process of integrating the input data from a set of probabilistic models known today in machine learning as the “expert result” [product of experts].

The meeting struck Freestone. He was inspired by the ideas of Hinton, and within the framework of intellectual cooperation sent him a set of his notes about his idea of ​​combining the seemingly “unrelated anatomical, physiological and psychological properties of the brain.” Freeston published these notes in 2005 - this was the first of dozens of papers in which he will develop his principle of free energy.

Blanket with a portraitAndrei Markov in the office of Carl Freestone: "keeps your inner states warm since 1856"

Also popular now: