Where are we up to in this series of Blogs?
The challenge for Artificial Intelligence is that we do not yet understand what consciousness is. Since ‘intelligence’ is a function of consciousness – it is one way in which we express consciousness – then this does not bode well. What it means is that AI can only ever be a function of our consciousness – we who build the robot. Anything that spells autonomous consciousness in a robot-child would have to pass through the evolutionary phases that humans passed through. Whatever consciousness is, it is itself a function of a long history and experience. Claims for AI to equip intelligent robots have to be somewhat reduced. The robot-child may learn (through trial-and-error, experimentation, etc.) but may never mature.
Now, there is a physical aspect to this – I mean, to the evolution of consciousness – and here, too, we see fundamental challenges for AI. Of course, the basis of all aspects of our humanity – physical, emotional, thought-based – is our genes – right down inside our DNA. We know an enormous amount about our genetic structure – we have mapped the genome. The genome is the structure of the 22,000 or so genes that humans have. We know each and every gene and we are learning just how they function. It may even be possible in a distant future to artificially build an autonomous genetic structure.
But this barely gets us started in building autonomous consciousness/ intelligence. Why?
Humans, surprisingly, share roughly the same number of genes as a banana and a bacterium. Our genotype seems to be merely what it takes to make for life. What life is and becomes – our phenotype – is a far more complex matter. This takes us into the realm of modern genetic research – epigenetics. I’m much less than an amateur at this – but here goes!
There is a level of cellular activity that goes on between the gene and the protein – the protein is what makes up living matter – skin, bone, etc. This concerns, not DNA, but RNA – different kinds of RNA are responsible for switching on and off (the term is ‘expressing’ or ‘suppressing’) configurations of genes in order to make the proteins we need. It is in this intermediate area – far more complex than the DNA (double-helix) structure in both its make-up and its workings – that what we are like is determined: the colour of our eyes, whether we are male or female (or ‘other’), our stress tolerances, our resistance to ageing, and so on.
Until recently, it was thought that Darwin’s view that animals evolve through exploiting mutations (genetic mistakes) that turn out to be advantageous and useful for survival. Genetic mistakes (Darwin did not know about genetic structures) are passed on through reproduction and those mistakes become ‘normalised’ in who we are. A giraffe had an errant gene that stretched its neck, this proved useful in reaching juicier leaves and that mistake was passed on to its young. Those who received the mistaken gene had a minor advantage in life and so their genes became dominant. A competing theory of Jean-Baptiste Lamarck said that animals develop certain characteristics in their lifetime in order to respond to the environment – leaves are high up so the giraffe develops a long neck and passes that on through inheritance to their young. Lamarck has almost universally been said to have been disproved. It’s nature, not nurture that accounts for who we are – it’s genetic evolution not circumstances that matter.
This provides for a level of stability – because evolution moves painfully slowly – which is useful for AI. If we could build a genetic structure, we could adapt it to ‘evolve’ consciousness.
But things have changed. It seems that Lamarck was – unwittingly – partially correct. Here’s a story.
Women were studied who had suffered severe hunger in their first three months of pregnancy. Their children – and even grandchildren – had a tendency towards obesity. It seems that, whatever the mother’s genes said about body mass, epigenetically the body had adapted the way genes were expressed so as to make optimum use of the few nutrients that came in. That characteristic – early in the development of the foetus – was inherited by the child whose body greedily squeezed every last calorie out of food. The genes did not change, just the way the genes were ‘expressed’ – and it was the expression that was passed on. Evolution was squeezed into a single generation in a Lamarckian way.
What this means is that we are not so stable as was thought before epigenetics burst into the scene. Our genetic make-up is more plastic – slightly more plastic – than we thought. This, indeed, is an important aspect of who we are – physically and mentally. Tendency towards anxiety can be inherited in this way, too. It is also likely that intelligence (and we haven’t begun to open up that can of worms) might be affected epigenetically, too – not in the crude and simplistic way that says breast milk produces more intelligent children than formula. But in terms of how the human animal responds to the immediate environment at times of extreme stress and adapts its epigenetic mechanisms to enhance forms of response to the environment (which is all intelligence must be).
This means that to achieve intelligent autonomy, to succeed in responding to the environment, a robot-child has to have the equivalent of an epigenetic infrastructure – not so much DNA as the more complex RNA. If it cannot respond in a sophisticated way it cannot be said to be ‘intelligent’. Once again, it relies on our intelligence to ease a response. If we are many years away from artificially reproducing a genotype, we must be hundreds of years away from reproducing an epigenetic structure.
So, having partially addressed the question of what it takes to reproduce humanity (i.e. to be autonomous as a robot-child), we should consider what it takes to reproduce humanism. Humanism is one of the higher-order responses to our environment – that which responds to the demands of living together densely and in mutually beneficial/prejudicial ways. That is for the next Blog.