Mark Zuckerberg, CEO of Facebook and lately a rampant giver-away of money, has announced his New Year’s Resolution: to build a robot butler to run his house, family, and life. In a Facebook post (obviously) posted on 3 January, he outlined his intention to “build a simple AI to run my home and help me with my work”. The finished product will apparently be like “Jarvis in Iron Man”, an AI powered personal assistant.
This seems a lot less achievable than his previous resolutions, which have included meeting a new person every day and learning Mandarin. Robotics is gaining speed, but it’s still an imperfect art. Here are a few of the bridges Zuckerberg will need to cross in his pursuit of an AI assistant.
1. Image recognition isn’t that good
In his post, Zuckerberg says he wants to teach his butler “to let friends in by looking at their faces when they ring the doorbell”. There are various problems with this.
First, while face recognition technology is very much established – Facebook itself uses it to try and tag pictures, often mistaking your aunt for some guy called James – it isn’t always that good. That can be down to the quality of the picture, which would be overridden if Zuckerberg has a CCTV camera which guests must look directly into, but it can also be because of glasses, sunglasses, hats, or tricks of the light. Google’s autotag feature, meanwhile, had a nasty habit of tagging black people as “gorillas”.
The second problem: the robot may recognise the face, but how many people would you actually want let into your house without your say-so? Presumably, the virtual butler would act like a normal butler, and ask your guest to wait for you in some sort of lobby area – in which case, why not come down and let them in yourself?
2. Robot motor function isn’t up to much either
Robots are now capable of some cogent conversation, but they’re still pretty poor at doing things with their “hands” outside a controlled or factory setting. The cereal robot above is a good example of this.
It therefore seems unlikely that Zuckerberg’s butler will have a physical presence, making it less useful for jobs like cleaning, or even preventing the rogue guests it let in from wandering into private rooms.
3. The internet of things is also (generally) terrible
Zuckerberg’s post explains that he’ll start with technology that’s “already out there”, implying he’ll draw on the internet of things – inanimate objects connected to the internet so they can operate in a “smart” way. But as enumerated daily by the Twitter account @InternetofShit and news stories reporting kettles leaking wifi passwords or TVs which record all your conversations, the internet of things is still very much in beta mode – if it takes off at all.
At the moment, the industry is beset by stories of easily-hacked networks, non-secure connections and products that just don’t do the basic things you want them to. Voice recognition, which many wifi connected devices rely on, can both turn on when it isn’t wanted, and, as in the case of this Samsung TV, store your voice to be listened to by anyone who gets access. Consumers are often left feeling that maybe they should have just stood up to boil the kettle themselves.
4. Would you trust an AI with your kids?
The butler will be responsible, apparently, for letting Zuckerberg or his wife know if anything is going on in his daughter’s room that he “should be aware of”. It seems unlikely that he’d trust the robot to make any crucial decisions on whether the improbably named Maxima is all right, which makes it seem like this particular feature will be, essentially, a baby monitor that will alert him to any sounds or movement in the baby’s room. Which is definitely already a thing.
5. Would it be a man or a woman?
As more and more robots are built, it’s becoming increasingly clear that they replicate the prejudices and biases we already hold. On the most basic level, we create robots to look and act like ourselves. But there’s also a disturbing trend towards female sex robots, self-service voices and receptionists, while we make male robots that do physically demanding tasks or complete mental activities. It would be reasonable to predict that the first robot to pass the Turing test will be given a male voice.
Were Zuckerberg to designate his robot slave/nanny a woman, he would be playing into this existing societal division. As noted by my colleague Helen Lewis in this week’s Sunday Times (£):
A third of the jobs in this country are at high risk of computerisation in the next 20 years, according to research from Oxford University. Respected economists are warning that robots will do to middle-class jobs this century what they did to manual ones in the last century.
Ultimately, this is why subservient female robot voices are such a problem. They cement the old stereotypes about which jobs should be held by men and which by women.
6. Why does he want a sentient assistant in the first place?
All in, hiring a staff member – or several – to do the work of the AI would be quicker and almost certainly cheaper than designing your own robot. Zuckerberg clearly sees this as an intellectual puzzle, and speaks in his post about the satisfaction of coding something for himself, rather than working with his developers.
Perhaps he has a god complex – after all Zuckerberg is one of the few people on earth with the wealth, resources and expertise to create his own personal robot. Did he just watch Ex Machina on Netflix and see himself in Oscar Isaac’s power-crazed search engine billionaire, founder of “Blue Book”, who creates a sentient AI? Stranger things have happened.