2 reasons you need to experiment with AI right now

Which statement do you believe? Robots will wipe out our jobs. AI and robotics will make everything free. These extreme viewpoints are both vying for our attention.

Singularity University, which aims to solve our global grand challenges through exponential technologies, widely reports that AI is the world’s cure. Peter Diamandis, co-founder and chairman of the university, says “rapid demonetisation of the cost of living” is earning his attention. Powered by technologies, he says, the cost of housing, healthcare, education and more will fall, “eventually approaching, believe it or not, zero”.

It’s easy to sensationalise superintelligence when you mix lofty institutional beliefs such as Singularity with fear from workers about the impact machines will have on their jobs.

Government ministers are right to show concern about the rise of unemployment and how the hollowing out of work might later affect skilled workers. But much talk embellishes and focuses on a future we can’t yet imagine. This is very early days – to use a cricket analogy, we're at lunchtime on the first day of a five-day match and we have a long way to go.

Where are we now?

There are many examples of firms chasing the next big idea. The latest is that machines can learn our most human traits. But it’s important to consider the counterargument.

Cogito is one such envelope-pushing firm. It’s an MIT spinoff that’s getting close to being able to detect your emotional state just by listening to your voice. Cogito is testing an app, Companion, which picks up on vocal cues that signal changes in mood and uses your phone to detect how active and happy you are.

Widespread use of Amazon’s voice assistant, Alexa, is proof that we’re comfortable integrating robo-voices into our lives, but are we ready for humanised technology? Let’s examine New Zealand-based start-up, Soul Machines, which develops emotionally responsive avatars. The “digital humans” are designed around the physiological model of a person’s face and have personalities to match the task they’ve been given. The inventors bring the avatars to life with neural networks that map the human brain.

Here’s the sticking point: toddlers are smarter than many of these computers. Gary Marcus, a professor of psychology at New York University and the founder of Geometric Intelligence, offers an insightful view.

He reminds us that, among other things, machines need thousands of examples to learn. Computers are great with big data, but they lack common sense. They make statistical observations about the way the world works. They fall down because they don’t ask “Why?” and have no underlying causal model – the way we, as humans, do things.

Researchers attempting to develop machines capable of engaging in natural conversation have more work to do. Some machines are already augmenting our work – for instance, sales call centres with bot advisors helping sales people land deals – but these outliers rely on tons of data from countless transcripts of real-life conversations. When it comes to cognitive science, that’s not quite how the human mind works.

Toddlers learn by extrapolating information. Their brains can recognise patterns in big data, but they can also use tiny pieces of information to make smart assumptions. For instance, if a child hurt their hand on an oven just once, they wouldn’t do it again. A self-driving car, on the other hand, needs thousands of miles of new road conditions to learn how to travel along it.

Intelligence involves perception, planning, analogy, reasoning, common sense and language. AI has made progress in the area of perception but it can’t beat us on empathy, creativity and relationships between things.

Much talk has centred on machines destroying work, but for many, machines will help knowledge workers do better.

Why get to grips with AI now

1. Free people up