Humanoid robots pose an ethical dilemma we've long prepared for

The National

2 days ago

Earlier this week, an Edinburgh University lab exemplified the second option. Looking like a toad made from rubbery stickle bricks, it's a 'soft robot' – one that can (with a whiff of air pressure) walk out of its own 3D-manufacturing unit. They'll be useful for nuclear decommissioning, biomedicine and space exploration, says the lab.
Great! Robots as curiously shaped facilitators of a cleaner, healthier, more ambitious world. Safely in the background.
READ MORE: I've voiced ScotRail trains for 20 years and was replaced with AI without being told
And then there are the humanoid robots (or HRs), currently cavorting all over your news feed. They're landing punches as Thai boxers in Hangzhou, China. They're playing badminton (admittedly with an extra two legs) in Zurich, Switzerland. A BMW factory, in the improbably named American town of Spartanburg, already has humanoid bipedal robots assembling parts on the production line (they're also starting in a Hyundai plant later this year).
Chinese state-run warehouses in Shanghai have human operators manipulating HRs, getting them to fold T-shirts, make sandwiches and open doors, over and over again. All generating data they can learn from, to act effectively in the near future.
Those who keep half an eye on radical technology may be a bit perplexed. Wasn't there some relief in the utter klutziness of robots, as they attempted to negotiate a few stairs, or turn a door knob? Didn't we share their pratfalls gleefully on social media – the bathos (if not hubris) that kept us relatively sane, in these accelerating times?
Computers might thrash us at most cognitive tasks. But tying shoelaces, making pizza, wiping a child's nose? Not yet, and maybe not ever. Hail the embodied human, and their evolved physical capabilities!
Well, there's a different track opening up. It's partly driven by the sci-fi imagination of the tech bros, East and West: most of these humanoid robots look like the rebellious droids in the 2004 movie I, Robot.
But it's also an assumption that the new, actively-learning models of artificial intelligence can do for humanoid robots what they've done for language, visuals and coding. Which is to generate plausible and coherent behaviour in the physical world, as they generate the same for prose or images.
Just to state the obvious: there's economic interest here. The target of these companies is a unit that can learn skills as required, flipping from task to task like a human worker.
'You can imagine a supply store has one, and that robot can be in the backroom depalletizing, cleaning, stocking shelves, checking inventory, just a huge range of things,' says Jonathan Hurst, co-founder and chief robot officer at Agility Robotics.
Working 24/7, only stopping to be charged: 'That's where the real value comes in', concludes Hurst.
READ MORE: Union slams Scottish companies using voice data without consent for new AI announcers
The point of a humanoid-like machine seems obvious to most of these entrepreneurs. The world is already designed for humans, and maximum profitability will come from robots stepping competently and confidently into this environment.
As the big business consultancies are predicting, the price for a working humanoid might descend to as low as $15,000 within the next few years, certainly lower as production scales up. That starts to become a viable business case for many enterprises – if the devices deliver on the performance promises currently being made.
Let's assume (and it may be a major assumption) that physical robotics is on the same exponential curve as the computations of AI (and indeed directly rides on the latter's ascent).
What that instantly opens up is a vast archive of myths and tropes about the fearful prospect of creating artificial humans, and how they'll live among us. We've been preparing for this ...
READ MORE: From the editor: We're all sick of Farage. But we can't 'just ignore him'
Greek mythology had Hephaestus making automatons – self-moving golden handmaidens
with 'intelligence in their hearts'. Pygmalion the sculptor fell in love with his statue Galatea; Talos, a giant bronze automaton powered by ichor, guarded Crete.
Across ancient India and China, still more defensive robots were imagined: the Buddhist text of Lokapannatti describes mechanical warriors that protected relics in subterranean cities. Back in Europe, the golem was raised to defend the Jewish ghettos. Mary Shelley's Frankenstein directly addresses our capacity to make humanoid subjects, and the ethics of the life we might share with them.
And we mustn't forget that the term 'robot' itself comes from the Czech genius Karel Čapek, and his 1920 play R.U.R. (Rossum's Universal Robots) – 'robota' being Czech for forced labour.
So right at the core of this domain's name sits the main anxiety we have about it. What does it mean for us to create an entity that we intend will work (or fight) entirely on our behalf?
It's one of the deeper, more civilisational arguments against a humanoid robot. Which is that it revives a master-slave framing from the worst of our past. An original cruelty of power that generates many others.
The great auteur of human weirdness, filmmaker Stanley Kubrick, articulated this well in his final movie, A.I. (posthumously realised by Steven Spielberg).
The robot boy David – and we shouldn't forget the underlying Pinnochio reference – is eventually discarded: he was a substitute for a real boy, who eventually revives from his coma.
Kubrick/Spielberg show how distorted human relations become – how resentful, harsh and violent – when these ever-more-perfect humanoid entities move among them. Their various roles of servitude do not protect them.
The end of the movie delivers a severe judgement on human morality. The robot boy is rediscovered, by beautifully communal 'mechas', on an utterly drowned and terminated Earth. On David's request, the mechs revive a clone of the human 'mother' who pushed him out into the forest. They are able to share one last, gentle day together.
READ MORE: How much has your MP claimed in expenses? See the full Scottish list here
The movie never fails to break my heart. But given the ethical dilemma it presents – do we really want to be masters in a society of slaves, again? – we might hope that the humanoids keep failing to turn that doorknob.
We really don't know what's coming. From the AI side, will developments in computation generate artificial consciousness, as well as artificial intelligence? By consciousness, I mean an entity which knows that it exists, experiences the world, has goals and desires and values.
If that intentionality and interiority appears, and begins to talk with us, we may anticipate one of its explicit interests: the rights of robots, operating under conditions of servitude. How might we respond? I'd suggest A.I. the movie shows us how we shouldn't.
Perhaps I'm operating in a very Western framework.
As many scholars note, animist religions (whether Japanese or Chinese) do not sharply divide between the animate and inanimate. The robot in Japan is not necessarily monstrous, but can possess 'kokoro' (heart/mind).
Osamu Tezuka's Astro Boy, flying about the comic strips of a traumatised Japan in the 1950s, was a great example of this. A moral child-robot with atomic powers, seeking justice.
I duly note that the most-watched Netflix production this week is The Wild Robot. The machine ROZZUM (Unit 7134) lands on an island teeming with wildlife, to which it slowly begins to relate and co-exist with. Kept in a bubble from marauding, egoistic humans, Roz is able to establish a kinship with these fundamentally different entities, evoking the most profound ecological themes.
Our sense of kinship with non-human animals should be obvious: the bass note of our responsibility to protect and honour the natural world. But should we prepare for kinship with these artificial entities? And should we ask whether casting them in humanoid form lays in more trouble than it's worth?
Edinburgh's stickle brick frog is made from gel, wobbly but ready for its limited tasks. It may be a more tractable robot than the gleaming Optimals marching – or we may still hope, shuffling – out of Silicon Valley.