12 June 2008

The Post-Humans Among Us

The recent IEEE Spectrum special on the singularity brought a lot of comment across the blogosphere. The NYT's John Tierney weighed in on the topic recently with a piece: When Do Post-Humans Show Up? Tierney gives Ray Kurzweil a chance to fight back against the "singularity deniers", and Kurzweil obliges.
These critics obviously have not read my book and have not read this chapter because they do not respond to anything I’ve written. It is as if they’ve just heard a superficial presentation of these ideas and respond without any engagement of the extensive discussion that has already taken place about these issues. _NYT
That may very well be, or it may be that Kurzweil is mentally fixed on a particular set of mechanisms and scenarios of singularity. It may be that Kurzweil's "extensive discussion that has already taken place about these issues..." is not as extensive or profound as Kurzweil imagines.

Kurzweil's discussion about how easily the human brain/mind will be emulated is particularly naive. This naivete comes naturally when a prolific and esteemed person such as Kurzweil is insufficiently familiar with the subject matter he is discussing--the genetics (and epigenetics) of the mind/brain.
I point out that the complexity of the design of the brain is at least 100 million times simpler than it appears because the design is in the genome. Even including the genetic machinery that implements the genome, the compressed genome is only about 50 million bytes (which I analyze in the book), and that is a level of complexity we can handle. We are already showing that we can develop realistic models and simulations of brain regions like the cerebellum and others. The cerebellum, for example, repeats a basic pattern a few billion times with some random variation within certain prescribed constraints. There is a lot of apparent complexity in the cerebellum but not very much unique design information, and we’re showing we can reverse-engineer it.
Of course the cerebellum is only peripherally involved in most conscious activity. It is an important "co-processor", but not the central processing center of consciousness. One can derive no comfort in the quest to understand human cognition from the apparent simplicity of cerebellar structure.

Similarly, if one supposed that the apparent simplicity of the brain genome implied a simplicity of the brain/mind itself, one would have to overlook much recent research detailing the "post-genomic", meta-genomic, and epigenetic development of central nervous system structures. These critical aspects of brain development are not well enough understood to allow useful modeling or quantification. Worse yet, even a complete understanding of how to create a human brain will not immediately put us in a place to understand how that brain works, or how it might be improved.

The road to the "singularity" will not be a smooth exponential curve. It will be a fractal fracturing of boundaries and limitations that will take decades to sort out. We will have pieces of the singularity existing a hand's breadth away from other pieces, with neither recognising the other. It will be up to post-humans to put the pieces together so that they do not blow up into a Skynet or Colossus.

If western civilisation survives attacks from desert religious fanaticisms, and 19th century cloistered ghetto-inspired central planning, various critical parts of the "singularity" may achieve capabilities and versatilities that allow them to connect with other critical parts in the same place at the same time. It is up to the post-humans among us to follow the threads of accomplishment, splice them together into a self-generative, autopoietic symbiotic whole, and wrap it all in a sustainable energy/matter matrix.

In Kurzweil's vision, the singularity drives the post-human. But doesn't it make more sense the other way around?

Eventually, the biological substrate of consciousness will be outpaced by other forms of conscious cognition. Post-humans will build their world around that knowledge, so as not to be left behind. Currently, only science fiction provides the speculative power to imagine the transformations that will come from genomics, nanotechnology, advanced hyper-parallel computation, robotics, evolved machine intelligences, and any combination of the above. After science fiction, Kurzweil provides a more "connected" view of our potential. Finally, there is mainstream science, which runs a very distant third in scope and vision to SF and Kurzweil.

But if you want a realistic assessment of what is likely to happen, you need scientist/engineers trained in multiple disciplines, who are also thoroughly steeped in biology, cognitive science, history, and science fiction. Post-humans will have to be able to bridge disciplines, cultures, even civilisations.

Labels: ,

Bookmark and Share


Blogger Barba Rija said...

I never thought of saying this Al Fin, but I have to say it:

I wholeheartedly agree with you.



(You must be thinking "where did I go wrong")

Friday, 13 June, 2008  
Blogger al fin said...

You know the old saying, "even a broken clock is correct two times a day?"

Friday, 13 June, 2008  
Blogger Snake Oil Baron said...

The statement about the brain being not nearly as complex as it seems is an interesting point. From the cellular level to the level of actual brain function, large amounts of the complexity are taken up doing organism-specific things that would not be necessary in an artificial system or brain "simulation".

At the level of brain function, it is an interesting exercise to wonder what a human level intelligence could get up to if it did not have to waste brain resources on the many organic (feeding and mating related) and societal (table manners and Christmas card lists) activities.

I just wish that we were further along in the endeavor. Artificial general intelligence has so many potential uses that even the initial production of the simplest commercialized forms would have vast repercussions.

Friday, 13 June, 2008  
Blogger al fin said...

Fascinating points, Baron. Remember, back in the 1950's, computer scientists believed they were right on the verge of creating intelligent machines.

Human consciousness is "embodied." We feel our thoughts in our bodies. Our thoughts make us hungry, thirsty, lusty, tired, aggressive, etc. You cannot separate our brains from our bodies, cognitively.

Machines will be different, when machines become sophisticated enough to become "conscious." But humans have trouble imagining how to make something like that--a type of consciousness that has never existed so far as we know. We have a hard enough time understanding our own cognition.

Jeff Hawkins has, in my opinion, taken the field the furthest so far--based upon a model of the human cerebral cortex. It makes sense. Start with what you know already works, and build on it.

All the machine intelligence groups who have attempted to work de novo, without regard to human/neural cognition, have gotten stuck in numerous dead ends and quagmires.

Fuzzy logic, neural nets, genetic algorithms, etc. are based on abstractions of the components of biological consciousness, at some level.

The internet and the potential for high level internet education, came along at just the right time to help humans escape from the "hyper-specialisation" trap that higher education has fallen into.

Current societal and technological roadblocks are caused--as often as not--by the inability of researchers and scholars to see outside their own fields.

Friday, 13 June, 2008  
Blogger Snake Oil Baron said...

Hawkins' book was fascinating but I find it hard to judge the progress his concepts have made with what little information I have seen about his new company.

I wonder if it will be as hard for a general artificial intelligence to be coordinated with more specific applications like database searching tools as it is trying to interface humans to such tools. The old style, non-intelligent computers have some powerful advantages yet they work so differently from how we believe intelligence works that making a system which can make the best use of both methods may be even harder than just designing general intelligence. It would be ironic if superhuman intelligences have to carry supercomputing slide rules in their... whatever they use for pockets.

Friday, 13 June, 2008  
Blogger al fin said...

Think modular. A superhuman intelligence would need to interface with co-processors and modules that performed various computational and storage/retrieval tasks.

The term "general artificial intelligence" can be misleading. More likely is a congregation of specialists with coordinator modules.

None of the modules alone would be intelligent or impressive.

Saturday, 14 June, 2008  

Post a Comment

“During times of universal deceit, telling the truth becomes a revolutionary act” _George Orwell

<< Home

Newer Posts Older Posts