drupal statistics module

Machines Like Us

Machines Like Us interviews: Susan Pockett

Monday, 20 June 2011

MLU: I think I'm in the "whole package" camp myself, for there are occasions when it feels to me that, indeed, my consciousness is just along for the ride. For instance, sometimes while watching a good movie I suddenly realize that I am smiling. In other words, I had been enjoying the film, but did not consciously realize it until becoming aware of the grin on my own face!

If consciousness is indeed a process and not a thing, do you believe that a sufficiently detailed model of the human brain -- an advanced computer simulation, for instance -- would be conscious? Searle, of course, says no, and Penrose feels that it is not even remotely possible.

SP: No, I don't believe a computer simulation could ever be conscious, or at least not anything that would run on computers as we currently build them. But that being said, I DO unequivocally believe that it will eventually be possible to generate consciousness using hardware instead of wetware. The trick will be to produce hardware that generates the right spatial electromagnetic patterns. This means that if you want consciousness (as opposed to intelligence), the whole connectionist paradigm is on completely the wrong track. Sure, artificial intelligence is possible using that means -- but if you want a conscious machine, the whole paradigm will have to be radically redefined.

MLU: If you received a multi-million dollar grand to build just such a sentient machine, in broad terms, how might it be designed? Where would you begin?

SP: Oh, what a delightful question! What would you do if you won the lottery? -- I love those discussions. Well, I think I would proceed on two fronts in parallel:

1. I'd put a bunch of engineers to work on the question of how to generate 3-dimensional electromagnetic patterns, on a sub-millimetre scale. I imagine the general approach would be some version of the mechanism that's used to produce spots of intense radiation in radiotherapy for cancer -- you aim two or more relatively low power beams so that they intersect at a particular spot, and at that spot there's constructive interference between the individual beams to produce a high-intensity field. But if I managed to hire clever enough people, perhaps one of them would come up with a whole new idea about how to do it.

2. I'd gather together another bunch of experts in neuroscience. Their overarching brief would be to determine the 3-D electromagnetic patterns that correlate with particular sensory experiences. I'd probably start with three groups -- one would work on auditory experience, one on somatosensory experience and one on visual experience -- and once a week they'd all have to meet and talk to each another, because it's very likely that there will be one basic pattern that goes with conscious sensations in general, with characteristic differences for each modality of sensation. Each of the three groups would consist of people who work on animals and people who work on humans -- neurosurgeons would be important parts of this second set. I'd start them off by suggesting a whole slew of specific experiments to get the ball rolling, none of which are currently part of any experimental paradigm that I'm aware of.

3. When uber-groups 1 and 2 had both made significant progress, I would probably open a third uber-group whose aim would be to translate the insights about what the wetware does that had been achieved by the neuroscientists into mechanisms that are easier to implement on hardware and interface with the products of uber-group 1.

So y'know, it's not going to be quick or easy -- or cheap. But I sincerely believe it's do-able.