Matthew Parris has narrated this article for you to listen to.
‘Et remarquant que cette vérité, je pense, donc je suis, était si ferme et si assurée, que toutes les plus extravagantes suppositions des Sceptiques n’étaient pas capables de l’ébranler, je jugeai que je pouvais la recevoir sans scrupule pour le premier principe de la Philosophie que je cherchais.’
Pardon my French – and I translate below. But so elemental was what René Descartes wrote (afterwards rendered in Latin ‘Cogito ergo sum’) that his phrasing should confront us first in his own language. Though in 1637 Descartes will have known nothing of robots, still less of artificial intelligence, he settled by this remark a debate that we think remains open, and which has recently resurfaced in media commentary. Can robots achieve consciousness? Might algorithms have feelings?
Last week my Times colleague, the science editor Tom Whipple, explored such questions in an article sub-headlined: ‘What if robots start to feel sad? What if that’s our fault? Academics say mankind must prepare for the possibility that AI machines may become sentient beings.’
Tom quotes Robert Long, executive director of the research organisation Eleos AI: ‘Evolution was not trying to build conscious creatures, it was trying to make creatures that would survive and reproduce. In the course of navigating the world and thinking about the world, that resulted in consciousness.’ What, he asks, if AI reaches the same destination?
A powerful argument. But it dodges a question almost every commentator is dodging: the biggest question. How will we know? Before we preach compassion towards machines, how will we know they can suffer? Before we learn to love or hate them, how will we know they can return our feelings?
Descartes’s answer is implicitly clear: we cannot know. By asserting that all we can prove is our own sentience, he implies the corollary: we cannot prove the sentience of others. But we presume. Should we extend the assumption to non-human intelligence?
This, in English, is what Descartes wrote: ‘And, observing that this truth, I think, therefore I am, was so certain and of such evidence that no ground of doubt, however extravagant, could be alleged by the Sceptics capable of shaking it, I concluded that I might, without scruple, accept it as the first principle of the philosophy of which I was in search.’ Though the Irish bishop George Berkeley later set out to criticise Descartes, Berkeley’s ‘immaterialism’ (the theory that only our own perceptions can be proved to exist, not what we perceive) is in fact complementary to Descartes’s dictum – as the 18th-century Scottish philosopher David Hume later remarked. Seeking proof of what exists, we are pushed back on to narrow ground: our own mind. Everything else, like Prospero’s island in Shakespeare’s Tempest, could be a dream, leaving only the existence of the dreamer as self-evident.
Seeking proof of what exists,
we are pushed back on to
narrow ground: our own mind
Bishop Berkeley tried to get round this by positing God as the master-projectionist. Descartes didn’t try. But I think the real force of Descartes’s argument has been overlooked by the centuries since his death. We have no proof of the existence of other minds. I cannot know you are sentient. I can know only that you display the outward signs of what within myself I experience as sentience. The outward signs, however, might be fabricated and the sentience illusory.
Indeed we know all about the fabrication of apparent sentience: we’ve been doing it for millennia with puppetry and more recently with talking dolls. Very, very recently we’ve been doing it with AI. We today strive mightily and with increasing success to design intelligence systems that give the appearance of consciousness and even of having empathy with us humans. From the British Gas chatbot to Siri, to Alexa, to the answers ChatGPT would offer if asked for a draft speech on human kindness, the aim of those who program these creations is to fashion a machine that seems almost human.
We’re doubtless only in the Stone Age of mastering this similitude. Yet even in the last century, cinema-goers shed an inward tear when in the Star Wars films the faithful robot manservant C-3PO appeared to die when his memory was erased.
C-3PO remained, however, under strict human control. Now (suggests Long) machines may break free and even without human programming may evolve to display consciousness and feeling, if these characteristics prove useful to their programmed purposes, as our own sentience and feeling have to our own race’s genetically programmed purposes.
Yes, but between the appearance and the reality lies a gulf that’s logically unbridgeable, and we will never cross it. As Descartes and Berkeley explained to our 17th- and 18th–century ancestors, we haven’t crossed it, even with other (presumed) humans. We cannot know whether their presumed sentience is real because we cannot by definition share what they are experiencing. We can have no window into another’s interior world. We know only that on encountering such situations as we too encounter, others behave more or less as we do, react and respond as we do. Knowing how we ourselves feel, we suppose they are feeling something similar. But it’s only a supposition: a supposition we’re programmed (by evolution) to proceed upon. And if we can’t do it for each other, we certainly can’t do it for machines.
My guess, however, is that as machines get more like us in their responses, we will gradually assume their interiors reflect their exteriors – just as ours do, and as a baby human gradually learns to do as it learns (or decides) that others must be copies of itself.
But we’ll never know. Descartes and Berkeley point the way. The question, reader, is not whether machines can feel. It is whether anything or anyone but you can feel. Watch The Truman Show. You too may be alone, but in a bubble designed to create the illusion that you are not alone.
Comments