Machine Love

 [As ever you can read this on the BBC News Website. It was inspired by a debate at the ICA which I chaired, and has attracted some comment online from Tara, Jordan,  and Clare, among others]

Much as I adore my MacBook I have no desire to form a life-long union with it or attempt to interface in any way that doesn’t involve keys, trackpad and my fingers.

Others seem to feel differently about the matter, like David Levy, who reckons that by the middle of the century our relationships with the machines that currently service our social lives will have grown significantly, and that intelligent robots will be sexual partners too.

He even thinks that human/machine marriages will be taking place, as people find the companionship they are looking for in ultra-realistic robots who never tire of them, never get bored with their jokes and never leave the toilet seat up. Mostly because they never go to the toilet, being robots.

Levy knows what he’s talking about, as an International Master who was closely involved in the development of chess-playing programs and a former winner of the Loebner Prize for programs that can have human-like conversations.

His recent book, ‘Love + Sex with Robots’, has attracted a lot of attention from computer scientists, psychologists and those just interested in the idea of having sex with a machine, so it was a real pleasure to have the chance to chair a discussion with him in front of a lively audience at London’s ICA.

In the end, however, I came away unconvinced.

The problem isn’t the humans, it’s the robots. Levy points out that artificial devices have been used for sex for many centuries, so there is nothing intrinsically implausible about the idea that humans may want to engage in sexual acts with robots.

He also argues that there is no fundamental reason why a human couldn’t fall in love with an artificial intelligence, and I think he’s probably right about this.

I don’t see why we shouldn’t form deep bonds with other conscious beings if they are capable of communicating with us, empathising with us, and being physically intimate with us.

Unfortunately this argument is entirely useless in practice because I can’t accept Levy’s sublimely optimistic view that we will be able to create such artificial intelligences within a 20-50 year time scale. In fact, I’m not sure we’ll ever manage it.

I don’t have any philosophical objection to the idea of ‘machine brains’, I just think that creating them will always be beyond our human capabilities. The AI research community has spent fifty years and billions of dollars to give us fragile systems that can play chess but do not understand why someone might ever want to do so. I believe that this will always be the case and that ‘strong AI’ with human-level intelligence or above will simply not happen.

And since Levy’s entire case depends on our ability to create these artificial consciousnesses and not merely superfast computers that can be programmed to simulate the full range of human emotions and interpret behaviour appropriately, the discussion about whether we will want to sleep with or marry these machines is entertaining but irrelevant.

It may be useful as a device with which to explore ethics, sexuality and human relationships, but we don’t need to worry about the practical implications for our legal system or religious leaders.

I am happy to project emotions onto my computer and see my laptop as trying hard to help me out when it struggles to render a video or run lots of programs at the same time. I will shout at the anti-virus software on my desktop when it kicks off a full system scan and slows the rest of the machine down to a crawl. And I’ve been known to plead silently with my phone to pick up a signal when I need to make an urgent call from a Welsh beach.

But this is anthropomorphism, assigning human qualities to machines that have neither soul nor intelligence, and even as I am doing it I am aware of what is happening. It keeps me happy, but I know it is not really going to change the outcome.

Just as we assign personalities to domestic pets despite their lack of consciousness, so we can attribute these qualities to the increasingly sophisticated robots that will soon be available in work and home settings.

And if it makes an elderly housebound person happier to imbue their house robot with human qualities, and if the programming can reflect apparent emotions in the robot’s behaviour that is fine. It may even help them live independently for longer.

But it is a long step from this to genuine engagement or emotion, and that is what we need if we’re going to have relationships with robots. It won’t be enough to have programmable partners whose characteristics we can define in advance, setting the degree of argumentativeness or affection to their preferred levels, or choosing a robot that sometimes behaves ‘unpredictably’ in order to add some excitement.

Chance and uncertainty are vital in any real human relationship. The lack of predictability, the contingency of love and the fear of rejection.  The fundamental asymmetry of a relationship between a human and a machine must surely debase it to the point where it could never be called ‘marriage’ and will just be another form of ownership.

If we want ‘real’ emotion we need ‘real’ intelligence.

I want my computers to be perfectly predictable because they are tools, extensions of my will. A good computer is a slave to my desires, a servant that listens and obeys my every whim, whether it is to write this particular sentence or close that particular file.

And I want my partners, whether for marriage or a night of passion, to have real feelings, real emotions and real needs and desires, not ones that have been put there to make me more easily tricked into intimacy.

The world of computing is littered with the broken promises of AI researchers who assumed that vision, hearing, movement and even consciousness were all achievable with a bit of programming and a faster computer. We would be well advised to treat the current collection of ‘imminent’ breakthroughs with a degree of scepticism.

Bill’s Links

Loebner Prize:
Economist on Japanese robots:
Levy book:
AI progress:
Arguing for strong AI: