Who's the cleverest one of all?

The future is here, and it’s not just talking back to us, it’s beating us at games. As artificial intelligence starts to develop intuition and creativity, should we be worried? Or are robots helping us far more than we know? Dave Waller sits down calmly to think things over

Chatbots

Illustration by Christian Montenegro

In March 2016 Lee Sedol, the South Korean world champion of the ancient Chinese board game Go, faced off against a computer program. He lost four out of their five games. In the course of its emphatic victory, the AlphaGo system, developed by Londonbased Google subsidiary DeepMind, used a move so striking that it’s been hailed as a key moment in the development of artificial intelligence, or AI. “It stunned Lee Sedol,” said DeepMind research scientist Thore Graepel. “It was the result of two neural networks working together and could represent the first signs of computational intuition and creativity.”

It’s precisely such intuition and creativity that links the innocuous act of winning a board game with the bigger stuff we’re used to seeing AI do in the movies. Recall, for example, HAL 9000, the talking computer in Stanley Kubrick’s 2001: A Space Odyssey. When HAL realises that its fellow space passengers are scheming to switch it off, it soon devises an ingenious method of protecting itself: destroying them. That’s the pattern when AI appears in fiction – things don’t tend to end well.

And such warnings haven’t exactly been confined to fantasy. In 2015, prominent thinkers Bill Gates, Stephen Hawking and Elon Musk of Tesla wrote an open letter cautioning against the unchecked development of the technology, arguing that mistakes made now may have unavoidable consequences down the line. It’s for this reason that Musk has called AI “our greatest existential threat”.

Whether AI’s ultimate impact is malign or not, one thing is clear: it has already moved into our daily lives in a huge range of subtle, specialised ways. It’s AI that trains Siri, the smart assistant that comes with Apple’s iPhone, to get better at voice recognition. It also enables your phone to identify callers not in your contacts, by spotting the number in an email you sent.

Netflix uses AI to work out your viewing habits and suggest the next thing you should watch; it even uses AI to gauge how much it should pay its film-makers. Emmanuel Mogenet, who heads Google’s major new AI research centre in Zurich, calls AI the ‘secret sauce’ – it’s found in Google’s search, spam filters and content removal, as well as the automatic transla - tion system it launched in September, and in Google Assistant, its rival to Siri. Facebook’s no different: AI enables it to recognise and tag faces in photos, and to decide what news you need to be reading.

AI is now a force in the workplace too. Milton Keynes-based Celaton describes inStream, its AI-powered software, as “the best knowledge worker you’ve ever hired”: the system analyses data your company receives, learns which bits you need and streamlines to make it useful. Elsewhere, AI protects businesses from cyber attacks: while traditional anti-virus software simply tries to block malicious intruders, Darktrace uses AI to monitor all activity across your networks, learn what’s normal and flag up anything suspect.

Many of these developments are happening thanks to a confluence of factors: computers are becoming more powerful, and breakthroughs in machine learning mean these more powerful computers get better at teaching themselves. Yet despite the progress, AI can still be clumsy. Facial recognition tech that enables the blurring of faces in Google’s Street View has been busy protecting the anonymity of statues and cattle; Amazon’s AI will conclude that because we just bought A Brief History of Time we now want to see its range of underpants and cuckoo clocks.

Amazon’s AI still concludes that, because we bought A Brief History of Time, we want to see its range of underpants and cuckoo clocks

The problem is that AI lacks regular common sense – a trait that happens to be very hard to program. Beyond its specialised tasks, AI still resembles a Dalek trying to go upstairs, or your friend who studied atomic computing at Cambridge, but can’t change a printer cartridge.

But things may be set to change. A key aim of Google’s new Zurich HQ is to build a ‘common-sense database’. “We are on the brink of a brand new era of computing,” Mogenet told reporters in June, adding that there was “no limit on how big I grow the team”.

Meanwhile, Google, Microsoft and Apple are racing to make their virtual assistants truly conversational, asking detailed questions of their users and teaching themselves to understand complex natural language. In May, Apple bought Cambridge-based AI developer VocalIQ, as its product was able to learn more than Siri from far fewer inputs.

VocalIQ’s team has spoken of its aim to build “an equivalent of KITT”, the brainy car from TV’s Knight Rider. We’re getting closer: AI-powered self-driving cars from Google subsidiary Waymo are due to hit the streets in 2017, Tesla’s fully automated cars will follow a year later. Uber is testing a selfdriving fleet in Pittsburg, while Honda is taking a different turn – unveiling its NeuV commuter cars, which will use AI “to artificially generate their own emotions”.

We may soon need all the emotional support we can get. If AI is fraught with ethical questions, none is more emotive than the risk the technology poses to human employment. Repetitive tasks are already falling to AI, but ‘safer’ fields, such as medicine and law, have been predicted to follow. “At some point, when the problem is not just Uber but driverless Uber, when radiologists are losing their jobs to AI, then we’re going to have to figure out how do we maintain a cohesive society and a cohesive democracy…” said Barack Obama, in an interview in November. AI needn’t spiral out of control before it asks some difficult questions of society.

The AI suffers no such crises. Having read twenty-five million medical studies in the first week of its new job, IBM’s Watson (itself a winner on TV quiz Jeopardy!) is analysing cancer patients’ medical records to help oncologists make more informed treatment decisions. AlphaGo remains the undisputed champion of Go. Meanwhile DeepMind’s next challenge is to get AI beating humans at StarCraft, a real-time war strategy video game. Now what can possibly go wrong there?



Buzz phrase:

Chatbots

What’s a chatbot, then, and how does it work?

A chatbot is a service, powered by AI, that you can interact with via a chat interface on your computer

Why do I need that?

It makes life so easy! Say you want to buy some shoes. You’d usually visit a website, then waste ages trawling through, hunting for the right style. With a chatbot, you simply explain what you’re after, and the chatbot shoots off to find it, or the closest option... Like visiting the store.

But chatbots do more than just sell shoes?

Much more. There are weather bots. News bots. Personal finance bots to help you manage your cash. Life advice bots to sort out your personal problems.

Ooh! So will I have to think at all?

Oh, the option’s there. Most certainly. If you really want to...

Contacting us is easy

Call us on 0345 279 8880

  • Monday to Friday08:30 - 17:30

You will be connected to a member of our team. They will be happy to help answer any questions you may have.

Let us call you back

Please complete as many fields as possible, the fields marked with * are required.