Artificial Intelligence, commonly referred to as ‘AI’ is the science and engineering of making intelligent machines.
Whilst an increasingly familiar term, its roots lie with Alan Turing. He led the team which famously cracked the ‘Enigma’ code machine used by Nazis during World War 2 to pass secret communications. In 1950, he created what has come to be called the ‘Turing Test’, which is still used as a means to determine whether a machine can effectively pass itself off as a human.
Anyone who remembers ‘HAL’ – the infamous AI from Stanley Kubrick’s ‘2001: A Space Odyssey’ from 1968 cannot fail to have been interested in the representation of AI portrayed in the film.
Despite a few hype cycles, often referred to as ‘AI winters’, expectations set by research and the reality it actually delivers are beginning to align. Aside from the behemoths like Apple, Facebook, Amazon and Google all ploughing vast sums into research, the landscape for startups continues to grow. In 2011, there were roughly 70 AI related startups which had achieved significant funding. That figure hit nearly 400 by end of 2015 and a further 200 had been added by June 2016.
It was widely reported in 2016 that DeepMind’s AlphaGo program (owned by Google) had beaten the 18-time world champion Lee Se-dol at the game ‘Go’. The significance of this cannot be underestimated as it was previously presumed a computer couldn’t beat a human champion due to the preconceived idea that Go is a game of intuition. While the rules of Go are straightforward, the game has more potential positions than atoms in the universe – and hence winning requires the ability to ‘think’ strategically. This victory is a step change in the way AI is perceived. However I’d argue it’s still nowhere near assuming human qualities. The machine wasn’t happy, had no one to share its £1m winnings with and just did what it was programmed to do.
Winning ‘Go’ is not the end game, of course. The big players recognise that products and services are moving away from ‘form’ or ‘app’ interfaces to a more conversational approach. Facebook’s ‘Deeptext’ is set to understand thousands of messages per second and in over 20 languages. Amazon’s deal with ‘Angel.ai’ in September 2016 opened the doors for it to automate customer service via Chatbots. Customers will be able to interact with the brand using natural language and get services automatically rather than requiring a human to make decisions.
Some of us are already using ‘Weak AI’ on our smart phones. Siri, Google Now and Cortana are all spoken-word digital assistants operating on iOS, Android and Windows Mobile. Research firm, Gartner, predicts that by 2020, 30% of all web-browsing sessions will be conducted without a screen, using voice search only. Another notable statistic is that by 2019, 20% of brands will abandon their mobile apps. AI will be the key facilitator as the step-change towards ‘chat’ based services at scale is predicated on it.
Moving into the physical realm, expect to see AI growth in areas such as self-driving cars. A key development is how the algorithms learn. In 2015, scientific journal ‘Nature’ published research demonstrating how Deepmind had created a program which was able to master Atari’s ‘Space Invaders’. Rather than being given a set of rules, it was simply given an objective – to achieve the highest score possible. Through repeated attempts, it achieved superhuman results. In other words – it won through experience. This matters for self-driving cars as, instead of programming maps, traffic lights or positions of obstacles, algorithms could play out billions of scenarios in a virtual environment, learning all the time and only then moving to physical tests. Certainly not self-aware, but learned from experience rather than pure programming. A key component of this ability to learn is the acquisition of huge amounts of real-world data; which is why Tesla, a leader in self-driving cars, sends thousands of parameters from every one of its vehicles back to base, to refine its algorithms automatically.
Despite all this, there are still serious weaknesses to be overcome. For example, if the onboard cameras are blinded by sunlight, they could fail to spot a changing traffic light. Sensors are still unable to recognise certain road obstacles. Humans find it easy to differentiate a lump of stone from a cardboard box. Machines don’t. The potential for a dangerous, unneeded swerve is absolutely still there.
How will AI influence our lives going forward? If we believe Science Fiction, not very well. Space Odyssey, The Matrix and Terminator, Ex-Machina all paint rather dystopian pictures. All fitting into two broad categories:
- Dominance
Robots control humanity – resulting in us being submissive or under threat of extinction. - Rebellion
Artificial Intelligence becomes self-aware and attempts to destroy us.
My sense is that we tend towards anthropomorphising. That is to say, we imbue AI with human characteristics – including the idea that a sentient AI will either love or hate us. This assumes that self-aware machines would have human value systems, which seems unlikely. There is however a mood amongst well known luminaries such as Stephen Hawking, Elon Musk (he of Tesla) and others that AI will doom us.
One thing’s for sure, AI will change the world. In the ‘Future of Work’ section of my forthcoming book, we already see the early impact of machines on jobs. Those jobs that can be automated will be automated, displacing millions of workers. Optimists love to proselytize how machines will create an abundance of near costless goods and services. In June 2016, we saw Switzerland conduct the first referendum on Universal Basic Income (UBI). Citizens were asked to vote on a proposal that everyone should receive a guaranteed basic income. Whilst it failed, supporters argue that as work is increasingly automated, fewer jobs are available. Expect to see much more about UBI as societies look for ways to deal with AI doing more of its work.
Looking further ahead, we see AI reaching a state of ‘Singularity’. Notable futurists such as Ray Kurzweil describe this as the point where exponential improvements in technologies such as genetics, nanotechnology, robotics, AI and computers reach a point where ‘progress is so rapid it outstrips humans’ ability to comprehend it’. According to Kurzweil, this will arrive in 2045, though this prediction is still widely disputed.
My sense is that we’ve reached a cross roads. In the short term, AI offers us fantastic opportunities for commerce and enrichening our lives through better products and services. Where it becomes murkier for me is that technological developments are largely uncontrolled and unpredictable. As systems become more interconnected, they have the potential to learn from one another. That’s where things get very uncertain. My natural inclination is to remain optimistic though. Look out for advances in disease elimination, surgical robots with precision control, the reduction of poverty and much more.