I recently interviewed author and inventor Ray Kurzweil about his new book, How to Create A Mind: The Secret of Human Thought Revealed. The 58-minute segment aired on December 1, 2 and 3 on the C-SPAN2 program After Words. The books thesis is that it is essentially possible to reverse-engineer the human brain to create a computer mind that works like yours and mine. The advantages of such a creation, Kurzweil told me, are three-fold. First, we can gain an improved understanding of the brain so that we are better able to fix problems with itfor example, developing new treatments for psychiatric and neurological disorders. Second, biologically inspired, more intelligent machines can help us solve numerous practical problems. Third, such a brain-replica may help us understand ourselves, and ultimately to help us become more intelligent. We are a human machine civilization and we create these tools to make ourselves smarter, he says. (Hear this response at about 13 minutes.)
These intelligent machines will take various forms, he suggests, evolving from the cell phones virtually all of us carry. If you ask a phone to do something for you, in most cases, it doesnt do what you ask on its own. The action, he says, takes place in the cloud. Future devices, he predicts, will also be gateways to the cloud. These gadgets will evolve in stages. They soon will be small enough to put in the display in your glasses as in Google Glass, which is a computer embedded in the frame of your glasses. Your screen, which displays all of visual reality, could be augmented to provide information so that when you look at someone, information about that person pops up. Just knowing a persons name will be a killer app, he quips. These systems will become intelligent, he predicts. Search engines wont wait to be asked for information; they will know you are struggling with something beforehand.
How will they know? They will be listening in on everything you are doing, everything you write, everything you read, everything you say, everything you hearif you let it, he says. The system, he says, should be listening in like a friend and realize: she needs this particular piece of information. Of course, no friend of mine can tap into my brain. And I have some serious reservations about uploading my thoughts to the cloud. But Kurzweil is optimistic about this future state of affairs. (Check out the dialogue starting at 17 minutes.)
Kurzweil predicts this technology will be blood cell-sized by the 2030s and 2040s, and will eventually live inside us. Tiny computers in our bloodstream will keep us healthy by augmenting our immune system. They will also enter our brains, enabling our neurons to act as gateways to the cloud and thereby expanding our mental capacities. (Listen to the chat at about 22 minutes.)
Schools for Computers
To create this future, Kurzweil says, you not only have to simulate the mammalian neocortex (the brains outermost layer), you have to educate it. If you dont, Kurzweil says, the resulting electronic brain wont do much. It might be cute, he suggests, but it will behave like a newborn. There are several approaches to educating these machines. For a detailed brain simulation such as the one being developed under the auspices of The Human Brain Project, you could use an electronic model as a teacher. That is, one computer could instruct another. When Kurzweil pioneered speech and character recognition systems in the 1970s and 1980s, he taught them by exposing the software to examples of real world speech. Watson, a computer system designed by IBM that beat two human champions at Jeopardy! in 2011, was educated over the course of weeks using Wikipedia and other encyclopedias. (Hear Kurzweils explanation at about 31 minutes.)
What about the body? Many researchers believe that our intellect is strongly influenced by interactions between our physical selves and the environment. I asked Kurzweil how we incorporate this concept, called embodied cognition, into our artificial brains if they lack eyes, limbs and the rest. Kurzweil acknowledges this limitation, but believes we have developed the capacity to conceive of abstract concepts in a way that is removed from the body. In addition, he suggests that there is a conceptual, or limited body for these artificial minds. Watson has the ability to speak, for example, to display text and to hear. In addition, the body can be a liability at times: We dont have to worry about the system having aggression or feeling jealousy, he points out. (This discussion starts at about 35 minutes.)
Humans have other deficiencies, too. Our memories are biased and distorted, for example, our decision-making suboptimal and our behavior often irrational. So I wondered whether our brain-based computers were really going to mimic such imperfections, or instead, perhaps improve upon us? Of course, he reminds me, we created computers in the first place to compensate for limitations of the human brain. A computer can remember literally, something we cant do. Critical thinking is another one of our weaknesses. Inconsistent notions often coexist in our minds. If you can identify and reconcile the inconsistencies, you can come out with some higher truth. Critical thinking is something we try to teach studentsand can similarly teach artificial brains, Kurzweil says. He proposes that our artificial minds include a critical thinking module that continually looks for, and tries to resolve, contradictions. (Tune into the conversation at 39 minutes.)
Kurzweils ideas about the human brain and artificial intelligence can be traced back to age five, when he first decided to become an inventor. As a young boy, Kurzweil had an inventory of parts from various construction toys hed been given and old electronic gadgets hed collected from neighbors. I had this idea that if I could figure out how to put these parts together in the right way, I could solve any problem, he recalls. At 12, he discovered the computer. Although most kids now discover the computer long before then, Kurzweil turned 12 in 1960, when, he estimated, only a dozen computers existed in all of New York City. He began programming, and wrote some statistical programs for the predecessor of Head Start. When Kurzweil was 14, he penned a paper about thinking. The paper proposed that the human brain was essentially a pattern recognizer, that it recognizes patterns really well, foreshadowing the ideas in his latest book. He then created a computer program that could find patterns in music and write original music using those patterns. The program won the Westinghouse Science Talent Search, among other contests. (Tune in at 52 minutes for this story.)
Ive been thinking about thinking for 50 years, Kurzweil says at the close of our conversation. Now, we have enough information to articulate a really clear theory of what I call the pattern recognition theory of mind and then use that biologically inspired paradigm to create even better AI.