Will we ever create an accurate simulation of the human mind? Can we detect and measure consciousness? When will artificial intelligence surpass human intelligence?
Humanity has questions about the development of AI, and for decades Ray Kurzweil has been trying to find the answers. Those who know the author, futurist, and inventor's work will be familiar with his beliefs in the exponential growth of information technology, and the inclusion of more technologies into the IT label. Lately, Kurzweil has become increasingly interested in the human mind, how we may be able to understand it, and eventually how we could recreate it. He's working on his seventh book, How the Mind Works and How to Build One, which will explore those concepts. This past August, at the annual Singularity Summit, Kurzweil gave attendees a sneak peak into his upcoming book via an hour long presentation with almost the the same name: "The Mind and How to Build One". Thanks to the Summit organizers, The Singularity Institute, Kurzweil's talk is now available to watch online; check it out in the video below. From his discussion on consciousness to his explanation of the processing methods of the cerebral cortex, this is one of the best Kurzweil presentations I've ever seen.
I attended this year's Singularity Summit, and had a great time. I remember some commenters at the Summit lamenting that Kurzweil started his talk rather slowly. However, I think the first 15 minutes of his presentation give some really valuable background to what he wants to discuss. Right away Kurzweil points out that the brain is not some mystic device, some quantumly unknowable system that we'll never be able to understand. We can, for the first time in history, reliably peer inside the brain and see what's happening. That's an important step in creating a comprehensive map of how our brain behaves. But in terms of AI, we may not really need that map. Kurzweil explains that reverse engineering the brain isn't absolutely necessary to develop artificial intelligence, it is just that understanding the brain can help us augment our pursuit of AI rather well. He relates how we've already had success with determining how the brain understands speech and visual input. These pattern recognition tasks have given us insight into how the rest of the organ processes information. With this context, Kurzweil's ready to jump into the future of creating artificial minds.
...But first he takes a bit of a detour. At 14:45 he starts to discuss the reasons why some people believe in the Singularity and others do not. Importantly, he points out that education, intelligence, and age aren't the determining factors. Glad to hear that the people who disagree with the concept of the Singularity aren't dumb, ignorant, or childish. At 17:44 he starts back towards the brain, explaining how the cerebral cortex is composed of modules, which he calls recognizers, that serve as linked labels for real world objects and metaphysical ideas. ...and then he gets away from the mind again. From 19:00 to 25:15 he shows evidence supporting the theory that information technologies have experienced exponential growth. For those who have seen Kurzweil speak before you can skip that part of the video. If this is your first Kurzweil presentation I have some bad news: Singularity Institute didn't include the slides in the video. Luckily I tracked down a similar presentation he gave to Google in 2009 (see it below). You can see all the graphs of exponential curves you'd ever want by checking out slides 5 through 44.
The real meat of the presentation starts up after 25:00 when Kurzweil really gets into exploring concepts related to the brain. Jump to that point in the video and you won't be disappointed. Slides 56 through 71 in the Google presentation are helpful to look through while you listen to him speak.
Unfortunately, Kurzweil was not able to appear in person for the Singularity Summit, instead he teleconferenced in. I was in the auditorium for the presentation, and I remember him looking a little like a giant floating head, but luckily you'll miss out on that when you see the video below.
**UPDATE 12.21.10 The video seems to have been taken down, we are actively working to resolve this issue.**
**UPDATE 12.22.10 The video is now back up! A better quality version, without the freezing halfway through, will be available soon. Thanks to Michael Annissimov and Singularity Institute for all the help.**
**UPDATE 1.10.10 A new video, without the freezes, is now available. It's included here.**
Here are the slides from Kurzweil's presentation at Google in July of 2009.
Part of why I like this presentation so much is that Kurzweil fills it with memorable statements that encourage the audience to learn more about the nature of their minds. At 26:04 he explains that consciousness, in its very nature, is not measurable. It is a subjective evaluation, not an objective one. Science is simply not going to be able to have a definitive test for consciousness. That's very appealing to me as both a challenge to experimentalists, and a launching point for philosophers. At 27:30, Kurzweil explains how thoughts create the brain saying, "we create who we are by the thoughts we have." Our thought patterns are literally rewiring our brain and our brain's wiring is influencing our thoughts. Speaking from experience, that's a wonderfully interesting concept to explore with friends over coffee late at night. At 39:25 he states that, "...the cerebral cortex is a LISP processor." Referencing the computer language LISP that uses linked lists as a data structure. Kurzweil describes the cortex as filled with units ("recognizers") that build complex concepts out of links to other concepts. That's a delightful (and apparently accurate) way to understand the way our minds learn, and again, something fun to discuss with friends or inspire you to read a book about neuroscience. It also jives very well with Jeff Hawkins' theories about the brain. Hawkins is the founder of Palm, Handspring, and most recently Numenta, a company that uses the architecture of the brain to help design narrow artificial intelligence for interesting things such as sorting through video footage. We here at the Hub are fans of Hawkins, and it's nice to see that apparently Kurzweil is too.
Further memorable sections:
43:00 - Kurweil discusses spindle neurons and the importance they have in our higher reasoning.
45:00 - He explains that we can only really test our perceptions of consciousness, not consciousness itself.
52:00 - The 'Duck Theory' of consciousness: If something looks like a duck, quacks like a duck, etc it's probably a duck. In the same way, humanity will likely decide to accept artificial entities as 'alive' when they do the things that our consciousnesses do, even if we don't have a test.
55:00 - Questions begin: 1) Is it possible the quantum wave function is a mental field? 2) How accurately do we need to model the brain to get intelligence? Neurons, subcellular, macromolecular? 3) Is scanning a human brain to the molecular level necessary before we get AI?
I should say that this presentation at the Singularity Summit has become a little frustrating to me. Around minute 30, Kurzweil starts to discuss the amount of code it would take to simulate a brain. A poor interpretation of these comments lead to PZ Myers, a researcher and blogger of some renown, to trash the entire presentation. We covered Myers' original blog posting, as well as Kurzweil's response, when it happened. As such I won't go into the debate too much here. Suffice to say that Kurzweil believes that our brains are encoded by our DNA, which represents a reasonable amount of code to try to simulate/recreate in the future. However, he also states outright that a simulated brain will need to be 'taught' because experience is a key element in the development of a mind (watch around 29:35). Myers seems to have missed all this and concluded that Kurzweil had a laughable naive comprehension of the complexity of the brain. Ugh. Misunderstandings such as these are not the best basis for reasonable debate.
Over the years Kurzweil's name has become somewhat synonymous with the Singularity. That's to be expected, I guess, since he has written so many books that have directly or indirectly discussed the topic. I often lament that equivalence because it opens up a complex intellectual concept to boring ad hominem counter arguments. Today, however, I'm rather glad that Kurzweil is so often portrayed as the leader of the Singularity. He doesn't always have the best stage presence, but it's hard to ignore the depth of thought and clarity of vision he brings to his presentations. At the Singularity Summit Kurzweil painted a detailed picture of the brain as we know it today, and the way we may delve it more deeply in the future. I look forward to reading his upcoming book to see how he expands upon these ideas.
[image credits: R. Kurzweil via Slideshare.net]
[screen capture and video credit: Singularity Institute]
[source: Singularity Summit 2010]