Robot Love

While I was in China, discussing the issue of consciousness in AI was central to the task I gave the students.  They made excellent videos to say that AI wasn't conscious, but was merely processing information and selecting words. "Merely" is the important word here. I said to them, after they had presented this, that I wasn't sure. A close encounter with a robot in my Chinese hotel has made me even less sure. Its not so much that the robot I met in the lift was "intelligent" (probably a lot less flexible in its choice of words than chatGPT), but that it could move around. Nothing which is conscious doesn't move. Everything conscious - from cells and bacteria (ok - that's a discussion) to bees and dolphins - moves to gain multiple perspectives on the environment. It was well-recognised as essential to perception by James Gibson in the 1950s. And here we are thinking that a legless computer program is or isn't conscious. Wait until it gets arms and legs.

The next question is much more complex but not unthinkable - might a robot love? The topic of love is so fundamental to the organisation of living things. It's also central to the plot of the Spielberg/Kubrick AI movie from 2001. How would we know if a robot loved? There are behavioural markers. Attachment is one - the behaviour that manifests in the maintaining of proximity to another system with which proximity is essential to balance of both systems. Will a robot react with depression and anxiety if an attachment is broken - if the source of stability is removed? Will a robot pine for the loss of its loved one? Will it grieve? Will it seek out a sense or a reminder of the loved one? Will it look for their scent, or will it trace their movements in vein hope of seeing them again? Will it compulsively check their inbox or an online forum for any sign that their loved one still thinks of them? And will they continue to harbour a longing in hope that one day they might be reunited? If it behaved like this, could it be said to have feelings? 

If it did behave like that, might it need (or benefit from) therapy? How would one engage a robot in therapy? I've got a PhD student looking at empathy and chatGPT at the moment - but its not just about the computer making speech acts that a therapist might. If the robot was the patient, what would be required to make it "feel" better? For us humans, perhaps we want to be "seen" and understood. Having someone in the position of a therapist say to us (for example) "You are mentally stable, etc..." matters. We might have known that anyway, and indeed we might have protested this to whoever suggested therapy in the first place, but it does matter for someone else to see it and say it. Perhaps it is fulfilment of curiosity about ourselves. For a robot to have feelings, it needs to be self-reflexive: to be curious about itself. Curiosity about ourselves can drive us mad, and other people can sometimes make things much worse. Good therapists can help stabilise our own self-inspection, maybe by reinforcing certain things we know to be true. 

Curiosity really matters. One could imagine a moving robot to encounter some phenomenon - either in the world, or within ourselves - about which it cannot make clear distinctions, and that in search of clarity, it would move to gain many different perspectives on that phenomenon - just as Gibson described. So an unfamiliar object would lead to some sort of compulsive behaviour - which perhaps one might associate with an object of love. We would at least learn something about our own emotional behaviour if this was possible: that so often the object of love is an object of fascination: loving relationships last because we remain ever-fascinated and wanting to learn more about loved ones. The feeling of love is the feeling of being on a journey. There is no reason why a robot should not go on that journey. 

I used to think that the chief weakness of our current AI is that it doesn't breathe. But then again, breathing is the result of millions of years of evolution and multicellular organisation. It isn't so much about gas exchange as about a sensitivity to the environmental conditions of the universe and the planet. And while the principal challenge of all living things is gas exchange, plants breathe in a very different way to the way we do. Why might a robot that becomes part of a multicellular network far more complex than any we can currently imagine, "breathe" in a completely different way to the way we do? Why might it not accommodate itself to the universe in a unique way?

Finally, I think the integration and speed of the combination of perception and anticipation is key to all of this, and this is what we are seeing in a crude form with current AI. I'm most interested in the application of this to music. My AI assistant can now transcribe the fine movements of my musical improvisation in codified forms which afford instant analysis and feedback. It is a simultaneous alternative description of the world, and the simultaneity of it really matters. Increases in speed are not just increases in speed. It is a fundamental change in the quality of something.