A Thousand Brains: A New Theory of Intelligence

by Jeff Hawkins

Unless we go out of our way to give machines equivalent fears and emotions, they will not care at all if they are shut down, disassembled, or scrapped.

At some point in the future, we will accept that any system that learns a model of the world, continuously remembers the states of that model, and recalls the remembered states will be conscious.

The recipe for designing an intelligent machine can be broken into three parts: embodiment, parts of the old brain, and the neocortex.

Human vision, touch, and hearing are achieved through arrays of sensors. For example, an eye is not a single sensor. It contains thousands of sensors arrayed on the back of the eye. Similarly, the body contains thousands of sensors arrayed on the skin. Intelligent machines will also have sensory arrays.

The important point is that for an intelligent machine to learn a model of the world, it needs sensory inputs that can be moved. Each individual sensor needs to be associated with a reference frame that tracks the location of the sensor relative to things in the world. There are many different types of sensors that an intelligent machine could possess. The best sensors for any particular application depend on what kind of world the machine exists in and what we hope the machine will learn.

The key idea is that intelligent machines will likely take many different forms. When we think about the future of machine intelligence and the implications it will have, we need to think broadly and not limit our ideas to the human and other animal forms that intelligence resides in today.

Behaviors that are intimately tied to the embodiment of a machine should be built in. For example, say we have a flying drone whose purpose is to deliver emergency supplies to people suffering from a natural disaster. We might make the drone intelligent, letting it assess on its own what areas are most in need and letting it coordinate with other drones when delivering its supplies. The “neocortex” of the drone cannot control all aspects of flight, nor would we want it to. The drone should have built-in behaviors for stable flight, landing, avoiding obstacles, etc. The intelligent part of the drone would not have to think about flight control in the same way that your neocortex does not have to think about balancing on two feet.

Finally, an intelligent machine must have goals and motivations. Human goals and motivations are complex. Some are driven by our genes, such as the desire for sex, food, and shelter. Emotions—such as fear, anger, and jealousy—can also have a large influence on how we behave. Some of our goals and motivations are more societal. For example, what is viewed as a successful life varies from culture to culture.

Some scientists, like me and my team, are theorists. We spend our days reading papers, debating possible theories, and writing software. Some of this work could, in principle, occur much faster if performed by an intelligent machine. But our software simulations would still take days to run. Plus, our theories are not developed in a vacuum; we are dependent upon experimental discoveries. The brain theory in this book was constrained and informed by the results from hundreds of experimental labs. Even if we were able to think a million times faster, we would still have to wait for the experimentalists to publish their results, and they cannot significantly speed up their experiments.

Once we fully understand what a column does and how to make one out of silicon, then it should be relatively easy to create intelligent machines of varying capacity by using more or fewer column elements.

There is one more way that we can think about capacity. Much of the volume of our brain is wiring, the axons and dendrites that connect neurons to each other. These are costly in terms of energy and space. To conserve energy, the brain is forced to limit the wiring and therefore limit what can be readily learned. When we are born, our neocortex has an overabundance of wiring. This is pared down significantly during the first few years of life. Presumably the brain is learning which connections are useful and which are not based on the early life experiences of the child.

Unlike the physical wiring in the brain, software allows all possible connections to be formed. This flexibility in connectivity could be one of the greatest advantages of machine intelligence over biological intelligence. It could allow intelligent machines to keep all their options open, as it removes one of the greatest barriers human adults face when trying to learn new things.

We start life knowing almost nothing and spend several decades learning. We go to school to learn, we read books to learn, and of course we learn via our personal experiences. Intelligent machines will also have to learn a model of the world. However, unlike humans, at any time we can make a copy of an intelligent machine, cloning it.

Because intelligent weapons can act without human supervision, they could be deployed by the tens of thousands. It is essential that we confront these threats and institute policies to prevent bad outcomes.

Notice again that there is no light, touch, or sound entering the brain. None of the perceptions that make up our mental experiences—from the fuzziness of a pet, to the sigh of a friend, to the colors of fall leaves—come through the sensory nerves. The nerves only send spikes. And since we do not perceive spikes, everything we do perceive must be fabricated in the brain. Even the most basic feelings of light, sound, and touch are creations of the brain; they only exist in its model of the world.

The two important points are that the brain only knows about a subset of the real world, and that what we perceive is our model of the world, not the world itself.

Our reality is similar to the brain-in-a-vat hypothesis; we live in a simulated world, but it is not in a computer—it is in our head.

These examples tell us that our model of the world can be incorrect. We can perceive things that don’t exist (such as the phantom limb), and we can incorrectly perceive things that do exist (such as the alien limb and the rubber hand). These are examples where the brain’s model is clearly wrong, and in a detrimental way. For example, phantom limb pain can be debilitating.

If you rely only on your personal experiences, then it is possible to live a fairly normal life and believe that the Earth is flat, that the moon landings were faked, that human activity is not changing the global climate, that species don’t evolve, that vaccines cause diseases, and that mass shootings are faked.

There is only one way, that we know of, to discern falsehoods from truths, one way to see if our model of the world has errors. That method is to actively seek evidence that contradicts our beliefs. Finding evidence that supports our beliefs is helpful, but not definitive. Finding contrary evidence, however, is proof that the model in our head is not right and needs to be modified. Actively looking for evidence to disprove our beliefs is the scientific method. It is the only approach we know of that can get us closer to the truth.

Today, in the beginning of the twenty-first century, false beliefs are rampant in the minds of billions of people. This is understandable for mysteries that have not yet been solved.

But what is disturbing to me is that billions of people still hold beliefs that have been proven false.

Because the neocortex is constantly making predictions to test its model of the world, the model is inherently self-correcting. On its own, a brain will inexorably move toward more and more accurate models of the world. But this process is thwarted, on a global scale, by viral false beliefs.

So, even though we are intelligent, our old brain is still here. It is still operating under the rules laid down by hundreds of millions of years of survival. We still fight for territory, we still fight for mating rights, and we still cheat, rape, and trick our fellow humans. Not everyone does these things, and we teach our children the behaviors we want them to exhibit, but a quick look at any day’s news will confirm that, as a species, across cultures and in every community, we haven’t yet been able to free ourselves from these less desirable primitive behaviors. Again, when I refer to a behavior as being less desirable, I mean from an individual or societal point of view. From the perspective of genes, all these behaviors are useful.

Today I have a good understanding of how the brain forms beliefs. In the previous chapter, I described how the brain’s model of the world can be inaccurate and why false beliefs can persist despite contrary evidence. For review, here are the three basic ingredients: 1. Cannot directly experience: False beliefs are almost always about things that we can’t directly experience. If we cannot observe something directly—if we can’t hear, touch, or see it ourselves—then we have to rely on what other people tell us. Who we listen to determines what we believe. 2. Ignore contrary evidence: To maintain a false belief, you have to dismiss evidence that contradicts it. Most false beliefs dictate behaviors and rationales for ignoring contrary evidence. 3. Viral spread: Viral false beliefs prescribe behaviors that encourage spreading the belief to other people.

Belief: Vaccinations Cause Autism 1. Cannot directly experience: No individual can directly sense whether vaccines cause autism; this requires a controlled study with many participants. 2. Ignore contrary evidence: You have to ignore the opinion of hundreds of scientists and medical personnel. Your rationale might be that these people are hiding the facts for personal gain or that they are ignorant of the truth. 3. Viral spread: You are told that by spreading this belief you are saving children from a debilitating condition. Therefore, you have a moral obligation to convince other people about the danger of vaccines.

Belief: Climate Change Is Not a Threat 1. Cannot directly experience: Global climate change is not something individual people can observe. Your local weather has always been variable, and there have always been extreme weather events. Looking out your window day to day, you cannot detect climate change. 2. Ignore contrary evidence: Policies to fight climate change harm the near-term interests of some people and their businesses. Multiple rationales are used to protect these interests, such as that climate scientists are making up data and creating scary scenarios just to get more funding, or that scientific studies are flawed. 3. Viral spread: Climate-change deniers claim that policies to mitigate climate change are an attempt to take away personal freedoms, perhaps to form a global government or benefit a political party. Therefore, to protect freedom and liberty, you have a moral obligation to convince others that climate change is not a threat.

Belief: There Is an Afterlife Belief in an afterlife has been around for a very long time. It seems to occupy a persistent niche in the world of false beliefs. 1. Cannot directly experience: No one can directly observe the afterlife. It is by nature unobservable. 2. Ignore contrary evidence: Unlike the other false beliefs, there are no scientific studies that show it isn’t true. Arguments against the existence of an afterlife are based mostly on lack of evidence. This makes it easier for believers to ignore claims that it does not exist. 3. Viral spread: Belief in the afterlife is viral. For example, a belief in heaven says your chance of going to heaven will increase if you try to convince others to believe in it as well.

Our old brain is highly adapted for short-term survival and for having as many offspring as possible. The old brain has its good side, such as nurturing our young and caring for friends and relatives. But it also has its bad side, such as antisocial behavior to garner resources and reproductive access, including murder and rape. Calling these “good” and “bad” is somewhat subjective. From a replicating gene’s point of view, they are all successful.

Now we find ourselves facing several existential threats. The first problem is that our old brain is still in charge and prevents us from making choices that support our long-term survival, such as reducing our population or eliminating nuclear weapons. The second problem is that the global technologies we have created are vulnerable to abuse by people with false beliefs. Just a few people with false beliefs can disrupt or misuse these technologies, such as by activating nuclear weapons. These people may believe that their actions are righteous and that they will be rewarded, perhaps in another life. Yet, the reality is that no such rewards would occur, while billions of people would suffer.

“Old Brain—New Brain” is the title of the first chapter in this book. It is also an underlying theme.

What makes us intelligent and self-aware? How did our species become intelligent? What is the destiny of intelligence and knowledge? I hope I have convinced you that not only are these questions answerable, but we are making excellent progress in answering them. I hope I have also convinced you that we should be concerned about the future of intelligence and knowledge, independent of our concern about the future of our species.