Jeff Hawkins is finally ready to explain his brain research.
  • Transfer

Jeff Hawkins, a Silicon Valley veteran who devoted the last decade to studying the mysteries of the human brain, organized a meeting with DeepMind, one of the leading AI laboratories in the world.

Scientists at Google’s parent company DeepMind, Alphabet Holding, want to create machines that can do everything the brain can do. Hawkins founded a small company with one goal - to find out how the brain works, and then recreate it, based on the knowledge gained.

The meeting scheduled for April at DeepMind in London did not take place. DeepMind has hundreds of AI researchers working together with a team of experienced neuroscientists. But when Hawkins talked before the visit with Demis Hassabis, one of the founders of DeepMind, they agreed that almost no one in the London laboratory would understand his work.

Hawkins says that before we can create artificial intelligence, we must explain human intelligence. Only then can we create machines that really work like a brain. “You don't need to reproduce the whole brain,” he said. "But it is necessary to understand exactly how it works, and to reproduce its most important parts."

In his company Numenta, this is exactly what he intends to do. Hawkins, 61, began his career as an engineer, created two classic computer companies, Palm and Handspring, and studied neurobiology along the way.

Now, after more than a decade of almost imperceptible work at Numenta, he believes that he and several researchers working with him are already on their way to solving the problem. At a conference in the Netherlands, he presented their latest research, which, according to him, explains the inner workings of the cortical columns - the main building block of brain functions.

How broad a community of researchers will react to Hawkins is difficult to predict: will they decide that the research should be studied more deeply? Or will they simply write Hawkins over the edge of self-assured and too unorthodox in their methods?

All this time, Hawkins followed his own, comprehensive idea of ​​how the brain works. This is a step beyond most neuroscience projects, be it the study of the brain of a fruit fly or the study of the characteristics of human vision.

The basis of his theory is cortical columns. They are an important part of the neocortex, the part of the brain that controls vision, hearing, language, and intelligence itself. Neurobiologists still have no consensus on how the neocortex works.

Hawkins is sure that the cortical columns treat each task in the same way, which is a kind of computer algorithm that repeats over and over. This is quite a logical approach to the brain for a person who has been building computing devices for decades.

All he needs is to figure out the algorithm.

Many neuroscientists like this idea, someone pursues a similar approach. Many praise Hawkins for his willingness to think so widely. Being an individualist in academia and the world of basic research is not easy. But a little easier if you can finance your own work, as Hawkins does.

However, some researchers are wondering if its self-financing activity, isolated from academic ties, is not a hobby? They examine the brain in small parts for a very reasonable reason - to understand how it all works together is a monumental, difficult task to reach.

“Obviously, we need a better understanding of the intellect,” said Tomaso Poggio, a neuroscientist at the Massachusetts Institute of Technology, who introduced the editors of Hawkins and Hassabis. "But Jeff chose a difficult path."

If the work of Hawkins succeeds, it can help researchers to go much further than what we see today. In recent years, companies such as Google, Apple and Amazon have created unmanned vehicles, gadgets that answer questions from different rooms, and applications for smartphones that instantly translate foreign languages.

They rely on neural networks, which are mathematical systems that are only to a certain extent similar to neurons in the brain. Scientists can not reproduce the brain, because they understand only certain parts of the picture of his work. Of course, they cannot simply take and copy its capabilities.

“The brain is undoubtedly the most difficult part of the matter of the universe we know, by any indicators,” said Christoph Koch, principal investigator and president of the Institute. Paul allen "We do not even understand the brain of the worm."

Call to explain the brain

In 1979, in an article in Scientific American magazine, Francis Crick, a Nobel Prize winner for DNA research, called for a comprehensive brain theory that could explain this "incredibly mysterious" organ.

Hawkins graduated from Cornell in 1979 with a degree in Electrical Engineering. Over the next several years, he worked at Intel and Grid Systems, one of the first notebook companies. But after reading the article in the journal, he decided that it was the brain that would become the work of his life.

He proposed the creation of a neuroscience lab at Intel. After the idea was rejected, he entered the University of California at Berkeley. The application for his doctoral thesis was also rejected. You could say he was an outcast.

In 1992, Hawkins founded Palm Computing. A decade and a half before the iPhone, he created a massive portable computer. When he hired an executive director, Donna Dubinsky, he immediately warned her that as soon as possible, he would quit his job at Palm and return to neuroscience. “It was always spinning in the background,” says Dubinsky.

US Robotics bought Palm in 1996 for $ 44 million. About two years later, Hawkins and Dubinsky left to establish Handspring. Palm, becoming independent again in 2000, acquired Handspring in 2003 for $ 192 million.

Around the second transaction, Hawkins established his own neuroscience laboratory. But it did not last long. He could not gather enough scientists focused on his cortical theory. So, together with Dubinsky and AI researcher, Dilip George, he founded the company Numenta.

Jeff Hawkins (left), Dilip George (center) and Donna Dubinsky. 2005 The

company spent years trying to create and sell software, but in the end, after George left, they focused on a single project. Funded mainly by Hawkins (he does not say how much he invested in it), the company's sole goal is to explain the neocortex and then reproduce it.

Inspiration coffee cup

In Numenta, Hawkins sits in a small office. Five other neuroscientists, mostly self-taught, work in another room behind the door.

Hawkins says that the moment of insight came two and a half years ago when he sat in his office and looked at the coffee cup.

He touched the cup and ran his finger along the rim. Then he jumped to his feet and ran out the door.

He ran headlong to his wife, who had just gone to dinner, and came across his closest associate, Subutay Ahmad, vice president of research. “Cortex knows the position of all things,” said Hawkins. Ahmad also had no idea what he was talking about.

When Hawkins looked at the cup, he realized that cortical columns not only perceive sensations, but also know the location of these sensations in space. They remember the world in three dimensions, and not in two. All the nuances are remembered in relation to what was around.

Hawkins, Subutai Ahmad and Donna Dubinsky. Office Numenta. 2018

Hawkins believes that if cortical columns treat visual and tactile information in this way, they perceive sound, language, and even mathematics in the same way. Since then, he has been working on the proof of this idea.

“When the brain builds a model of the world, everything is relative to everything else,” said Hawkins. "That's how he understands everything around."

The source of tension between Hawkins and other researchers is not at all in the fact that they believe that he is wrong. They just don't know it, because he is trying to do it in a completely different way. And so wildly ambitious.

For the further development of science, what Hawkins is working on cannot remain on the margins. His ideas can bear fruit, thanks to experiments with other neuroscientists, said Nelson Sprouston, senior director of the Janelia Research Campus, a research laboratory in Virginia that deals with neuroscience. “The continuous cycle of testing and revising the models of neural computing prompted by biology is key to the development of breakthrough brain theories,” he said.

To paraphrase, Hawkins needs to be given the opportunity to thoroughly research his work and find a way to interact with scientists who, most likely, have never looked at the brain like it.

At the moment, a translation of a great scientific article by Jemen Hawkins and his colleagues from Numenta is being prepared: “ A Learning the Learning about the Structure of the World .” Also the second big article “ A Framework for Intelligence and Cortical Function Based on Grid Cells in the Neocortex ” is planned for translation .

Also popular now: