21 December 2012

Does Kurzweil know How to Create a Mind?

Only if a "leap of faith" = knowing.

I read Ray Kurzweil's new book "How to Create a Mind - The Secret of Human Thought Revealed" (a title which has a pretty hucksterish tone, doesn't it?), and have a few thoughts.

As I've noticed before in reading him, he has a tendency to lump issues of different orders of magnitude together as if they are equal in complexity.  Here's one line had me shaking my head - after discussing plans for using a Watson-like system for helping to treat human disease, on page 108 he writes "Future systems can have goals such as actually curing disease and alleviating poverty."  This seems to indicate that alleviating poverty is just some new algorithm, ignoring completely issues of power and politics.

But as for the entire book, it's a bit of a mish-mash of his earlier stuff.  He spends quite a big portion of the book discussing his notions around the neuroscience of the neocortex, and the idea of pattern-matching modules which have a general architecture which helps allow for plasticity of brain function.  He does at least acknowledge that the Jeff Hawkins book "On Intelligence" covered some of this ground already (and I'd say in a more readable fashion).  I can't say whether this view of the brain is really backed up by all the latest research, but in general it seems plausible.

But when we get to the more interesting (to me at least) questions of mind and consciousness, Kurzweil doesn't really pin things down in any way.  Most of my attention was drawn to Chapter 9, "Thought Experiments on the Mind".  Kurzweil writes that his view is that "consciousness is an emergent property of a complex physical system" (203) - note that he uses "physical system", not a biological system, and follows up with his belief:  "By this reckoning, a computer that is successfully emulating the complexity of a human brain would also have the same emergent consciousness as a human."

The closest thing I could find to a definition of mind comes a few pages later; "I refer to 'mind' in the title of this book rather than 'brain' because a mind is a brain that is conscious." (205).  So here again we find the equivalence of mind to brain, and I find this an inadequate way of thinking about the issue. I believe that humans can be conscious, not brains, and I believe that 'mind' is about subjective experience, not about brains (while acknowledging that a healthy brain is a necessary part of the equation).  Whether a computer can equal a brain is not so clear either, though clearly we are making progress in modeling brain-type functions.

As for the 'leap of faith', Kurzweil uses the phrase many times.  "My own leap of faith is this: Once machines do succeed in being convincing when they speak of their own qualia and conscious experiences, they will indeed constitute conscious persons." (210)  I admit that there well may be machines in the future that are very convincing, but it really is a leap of faith to believe that they will have subjective experiences in some way like humans.  I believe there is still plenty to be learned about human consciousness, and its relationship to the brain, regardless of the evolution of machines.

Then he has a section on free will which seems to me to build upon some of the confusion of mind-consciousness-brain.  He discusses research on split-brain patients where the two hemispheres are not directly communicating, and where it seems clear that each hemisphere is capable of operating essentially independently.  From this he concludes: "This implies that each of the two hemispheres in a split-brain patient has its own consciousness." (227).  I don't believe that follows.  The question is whether the split-brain person experiences two separate consciousnesses, and I don't believe that they do give such a self report - and given that I don't see the brain (or part of a brain) itself as conscious.  On 233 he states "We consider human brains to be conscious" - but NO, I don't - it is humans themselves who are conscious.

Still in Chapter 9, Kurzweil provides a thought experiment on identity.  First he describes a scanned 'copy' of a person, which is loaded onto a non-biological platform, and seems to behave just as the original.  He claims, without much justification in my mind, that this copy is conscious, but also initially concludes that is separate from the original, even if very similar.  This latter conclusion I agree with.  I believe a copy may be very like the original, but as soon as the copy is made it is on a separate path with separate experiences (separate mind, if it has one), and thus cannot be equivalent (just as identical twins are not equivalent).

Then he describes a piecemeal 'replacement' program on a person, bit-by-bit replacing brain components with non-biological components, until finally at some point the brain is completely replaced.  In this case it seems as if identity is retained, and I believe that makes sense.  But he says there is a contradiction here which indicates that this replaced person is fully equivalent to the copy above, and thus the copy is (has the same identity as) the original.

I think there is a crucial difference in the two scenarios, and that is in the second scenario, identity is retained because there is a ongoing continuity of experience - indeed I think that is what is crucial about identity.    Kurzweil never really defines identity, just as he never really can define consciousness either, and I think this leads to all sorts of muddy thinking.  I'm not saying it's easy to define these terms, but one must make some stab at it or else the words stand for very fuzzy concepts.

In the final portions of the book Kurzweil goes over his Law of Increasing Returns (LOAR) that was in The Singularity is Near, dealing with some objections raised by Paul Allen among others.  Not much new here.

To sum it up, I think Kurzweil is right about the trends in terms of increasing hardware and software power (in smaller packages), and this will indeed lead to some impressive breakthroughs.  Perhaps this will include the creation of robots that seem very human.  Whether or not they have minds or consciousness is certainly not resolved in this book, and we may just have to wait and see.  I think his book would be better titled "How to Create a Non-biological Brain" and leave it there.


17 December 2012

Mind moves matter!

Or at least gets neurons to fire - paralyzed woman can control robotic arm.  The Guardian story "Mind over matter helps paralyzed woman control robotic arm" by Ian Sample on 16 December 2012 reports on developments in a Pittsburgh case.

The 52-year-old patient, called Jan, lost the use of her limbs more than 10 years ago to a degenerative disease that damaged her spinal cord. The disruption to her nervous system was the equivalent to having a broken neck.

...

Writing in the Lancet, researchers said Jan was able to move the robotic arm back, forward, right, left, and up and down only two days into her training. Within weeks she could reach out, and change the position of the hand to pick up objects on a table, including cones, blocks and small balls, and put them down at another location.
And here's the part I especially like:
To wire the woman up to the arm, doctors performed a four-hour operation to implant two tiny grids of electrodes, measuring 4mm on each side, into Jan's brain. Each grid has 96 little electrodes that stick out 1.5mm. The electrodes were pushed just beneath the surface of the brain, near neurons that control hand and arm movement in the motor cortex.

Once the surgeons had implanted the electrodes, they replaced the part of the skull they had removed to expose the brain. Wires from the electrodes ran to connectors on the patient's head, which doctors could then use to plug the patient into the computer system and robotic arm.

Before Jan could use the arm, doctors had to record her brain activity imagining various arm movements. To do this, they asked her to watch the robotic arm as it performed various moves, and got her to imagine moving her own arm in the same way.

While she was thinking, the computer recorded the electrical activity from individual neurons in her brain.
So "thinking" and "imagining" seem to have direct physical impacts in the brain.  What more can we learn to achieve by thinking?

05 December 2012

Neuroscience Fiction

Gary Marcus in the New Yorker on brain complexity, "Neuroscience Fiction", points out that our current tools like fMRI may not really show all the important things going on under the hood:
But a lot of those reports are based on a false premise: that neural tissue that lights up most in the brain is the only tissue involved in some cognitive function. The brain, though, rarely works that way. Most of the interesting things that the brain does involve many different pieces of tissue working together. Saying that emotion is in the amygdala, or that decision-making is the prefrontal cortex, is at best a shorthand, and a misleading one at that.
He also links to a recent NYTimes op-ed complaining about the recent prevalence of 'brain porn' by Aissa Quart:
A team of British scientists recently analyzed nearly 3,000 neuroscientific articles published in the British press between 2000 and 2010 and found that the media regularly distorts and embellishes the findings of scientific studies. Writing in the journal Neuron, the researchers concluded that “logically irrelevant neuroscience information imbues an argument with authoritative, scientific credibility.” Another way of saying this is that bogus science gives vague, undisciplined thinking the look of seriousness and truth.
Perhaps this is all just an inevitable backlash against the early peak of hype.  There's a long way to go with neuroscience and actual understanding.

04 December 2012

Human Connectome Project

Along similar lines to yesterday's post, the Human Connectome Project looks interesting, attempting to create models of neural pathways.  Here's an article, "The Symphony Inside Your Head" by Dr. Francis Collins that discusses the project; an excerpt:
For some time, neuroscientists have been able to infer loosely the main functions of certain brain regions by studying patients with head injuries, brain tumors, and neurological diseases—or by measuring levels of oxygen or glucose consumption in healthy people’s brains during particular activities. But all along it’s been rather clear that these inferences were overly simplistic.  Now, new advances in computer science, math, and imaging and data visualization are empowering us to study the human brain as an entire organ, and at a level of detail not previously imagined possible in a living person.
Worth a look!

03 December 2012

Brain Simulation

The Nature article "Simulated brain scores top test marks" by Ed Yong (Nov. 29, 2012) describes a project to simulate the neurons of a brain via computer.  It's called Spaun!
Spaun sees a series of digits: 1 2 3; 5 6 7; 3 4 ?. Its neurons fire, and it calculates the next logical number in the sequence. It scrawls out a 5, in legible if messy writing.

This is an unremarkable feat for a human, but Spaun is actually a simulated brain. It contains 2.5 million virtual neurons — many fewer than the 86 billion in the average human head, but enough to recognize lists of numbers, do simple arithmetic and solve reasoning problems.
While the current version is quite slow, this seems like an interesting development!