2015 : WHAT DO YOU THINK ABOUT MACHINES THAT THINK?

juan_enriquez's picture
Managing Director, Excel Venture Management; Co-author (with Steve Gullans), Evolving Ourselves
Head Transplants?

 

In the pantheon of gruesome medical experiments few match head transplants. Animal experiments have attempted this procedure in two ways: substitute one head for another or graft a second head onto an animal. So far the procedure has not been very successful. But we are getting far better at vascular surgery, bypassing, stitching, and grafting both big and microscopic vessels. Similar advances are taking place in rebuilding muscles and damaged vertebrae. Even the reattachment of severed spinal cords, in mice and primates, seems to be advancing steadily.

Partial brain transplants are likely a long way out. Other than some stem cell procedures, attaching parts of one brain to another is highly complex given the consistency of most brain mass and the trillions of connections. But as extreme operations, reattachments of fingers, limbs, even faces, become commonplace the question of whether we could, and should, transplant an entire human head loom closer.

Partly reattaching a human head is already a reality. In 2002 a drunk driver hit teenager Marcos Parra so hard Parra's head was almost entirely detached; only the spinal cord, and a few blood vessels, kept the entire cranium from coming off. Fortunately a surgeon, Curtis Dickman, had been preparing for just this type of emergency. Screws reattached vertebrae to the base of the skull, part of the pelvic bone was redeployed to bring neck and head back together, and within six months Parra was playing basketball.

Successful animal whole head transplants may not be that far out. And if such procedures were successful, and the animal regained consciousness, one could begin to answer pretty fundamental questions including: do the donor's memories and consciousness also transplant?

Similar questions were asked during the first heart transplants, but it turns out the emotions, attachments, and loves of the donor did not transplant with the organ that was always "tied" to emotions. The heart is but a muscle. How about the brain?

If mice with new heads recognized previously navigated mazes, or maintained the previous mouse's conditioned reactions to certain foods, smells, or stimuli, we would have to consider the possibility that memory and consciousness do transplant. But if experiment after experiment demonstrated no previous knowledge or emotions, then we would have to consider that the brain too might just be an electro chemical muscle.

Actually knowing if you can transplant knowledge and emotions from one body to another goes a long way towards answering the question "could we ever download and store part of our brains, not just into another body but eventually into a chip, into a machine?" If you could, then it would make the path to large scale AI far easier. We would simply have to copy, merge, and augment existing data, data that we would know is transferable, stackable, manipulatable. The remaining question would be: what is the most efficient interface between the biology and the machine.

But if it turned out that all data erases upon transplant, that knowledge is unique to the individual organism, (in other words that there is something innate and individual to consciousness-knowledge- intelligence), then simply copying the dazzlingly complex connectome of brains into machines would likely not lead to an operative intelligence.

If brain data is not transferable, or replicable, then developing AI would require building a parallel machine thought system, something quite separate and distinct from animal and human intelligence. Building consciousness from scratch implies following a new and very different evolutionary path to that of human intelligence. Likely this new system would eventually operate under very different rules and constraints. In which case it would likely be far better at certain tasks and be unable to emulate some forms of our intelligence. Were AI to emerge from this kind of evolutionary system it would likely represent a new, distinct consciousness, one on a parallel evolutionary track. In this scenario how machines might think, feel, govern could have little to do with the billions of years of animal-human intelligence and learning. Nor would they be constrained to organize their society, and its rules, as do we.