Review: The Hunt for AI

Watson plays Jeopardy!

Horizon: The Hunt for AI
BBC HD, 3 April 2012

As the stern grandfather of the BBC’s science output, Horizon is as topical as ever with its latest programme exploring the search for artificial intelligence. In reference to all things brainy, modern-day mathematician and Professor of Public Understanding of Science Marcus du Sautoy leads the viewer on a journey from Bletchley Park to Berlin and back via tightrope walking and stumbling robot babies.

The programme draws upon a range of agents to explore what really makes us human and if a robot can ever exhibit the same level of comprehension and interaction as we do. We are also challenged to think exactly what it is that defines intelligence – is it the ability to correctly answer questions in a foreign language, simply by deciphering a code and rearranging it? Or is intelligence ‘embodied’ – tied in with having a body that we can use to move and feel? The search for AI is certainly throwing up more questions than its pursuers imagined.

One of the robot stars of the programme is Watson, a supercomputer built by IBM to discover whether a robot could be programmed to outwit human competitors in the classic US gameshow Jeopardy! by answering the quiz’s fiendishly cryptic questions in natural human language. Waiting for his prodigy to take to the stage, David Ferrucci admitted feeling a sense of trepidation about how Watson would be received. “People are going to root for the people rather than the bloodless animatronic machine”, he says. This fear about the potential for AI science to run away with us could easily have formed a key strand of the programme, however Du Sautoy quickly deflects it with his own personal attempts to understand how to think in a non-human way.

Running through the hour-long documentary like a stuttering time-bar is Marcus’ attempts to walk a circus tightrope, a task he can apparently only complete if he stops thinking like a human and relies only on the instincts of his body. Unfortunately, this metaphor is less than successful, implying only that the brains of scientists are just too full of important knowledge to be switched off. Instead of contemplating the nature of learning a new skill, I was left thinking that perhaps middle-aged men are just not very good at circus tricks.

The high point of the programme came from Du Sautoy’s journey to visit a lab not only in the outskirts of Berlin but also on the outer edges of artificial intelligence itself. In a small, unremarkable room, two robots are learning like human toddlers: by looking at their reflections and communicating with each other. And they can interact with humans too (though I’m sceptical about the man on the laptop in the background). Du Sautoy’s reaction to the alarmingly cute machine almost falling over is a joy to watch: this is a man genuinely excited about the future of artificial intelligence.

But why should we bother searching for a way to replicate the tasks we are perfectly capable of carrying out by ourselves? In the words of Du Sautoy, “have we become obsessed by our own abilities?” I think not. What is the purpose of science if not to discover the limits of what we can achieve? I don’t think that we should be worried about an uprising by our own creations, I, Robot style. Oh, and by the way, Watson won.

Horizon: The Hunt for AI can be watched on BBC iPlayer until 17 April.

2 thoughts on “Review: The Hunt for AI

  1. I completely agree about the tightrope analogy. I just started to get quite annoyed, as he seemingly wasn’t listening to what he was being told either. I thought the rest of the show was very good though, and on the whole I liked Du Sautoy as the presenter.

  2. I actually disagree about the tightrope demonstration, as it’s not really an analogy at all – De Sautoy does improve at walking the tightrope, demonstrating a human’s ability to learn seemingly arbitrary tasks in a short timescale. Most AI research is not directed towards ‘general intelligence’ or ability; instead we focus on specific areas of expertise, and having De Sautoy actually complete a tightrope section after a day’s practice shows the difference between AI and humankind.

    I enjoyed parts of the documentary, but certain sections were poorly handled, and I think the Berlin piece especially so. Steels talks about his work in a very offhand way, announcing that it’s hard merely to make robots stand up. This is true, of course, but it’s also a problem that many people have solved and continue to solve in robotics labs around the world every day. Beyond that, beyond the challenge of carrying out a task in a situated environment, the robots are not doing anything particularly complex.

    And this was what undermined the programme for me. The question of embodiment is an interesting one, but it hardly defines the modern frontier of artificial intelligence. Watson and The Painting Fool were given less in-depth explorations compared to the robotic work, yet the tasks they accomplish are considerably more intricate and – I’d say, at least – intelligent. Yes, doing things using cameras and sensors is tricky, but it’s a small portion of the AI world, a world that is beginning to write poetry, design bridges, detect terrorists and save lives. Robots are cute – but they’re not the whole story.

Leave a Reply

Your email address will not be published. Required fields are marked *