AI will be seized upon by start-ups and pornographers who will make more ingenious use of it once it becomes affordable, but the lead will come from those with billions to invest
As a fundamentally childish person, when the topic turns to Artificial Intelligence my mind runs to C3PO first, and Skynet second. Inevitably when we think about the future of machine learning, the shadow of The Terminator looms large. All nonsense, of course – computers won’t wipe us out because of a lust for power. But they could do it by accident – as posited in a wonderful talk by Daniel Hulme of Satalia at the field.work data festival in London.
Although Hulme doesn’t think anything of the kind will happen, and is a huge advocate for better decisions and processes designed by humans and machines in tandem, he did point out that you have to pose the right questions. If you asked a computer the best way to eliminate a disease like cancer without framing the question in the right way – it might well suggest wiping out all the disease carriers. Which doesn’t mean it would go ahead and launch nuclear Armageddon through a modem. Hopefully.
Computer scientist and futurist Ray Kurzweil has predicted that ‘the singularity’ – the point at which cognitive computing power surpasses human intelligence – will arrive as early as 2029. Whatever the date – and the actual tipping point doesn’t matter so much – burgeoning Artificial Intelligence presents huge challenges in terms of employment, skills and governance, and huge opportunities in terms of scientific and social problem solving.
2029 is probably a bit soon. Anyone who has ever spent a week trying to teach an analytics package to recognise Boolean strings knows that computer learning is at once brilliant, and frustratingly clunky. Companies worldwide are trumpeting their ability to pull together buying behaviour, social and search data to spot trends and opportunities for marketing, but the outputs are more of a springboard than a solution. Cognitive data crunching can guide investment and assess the impact of creative, but fortunately for the skills economy these insights still need creative interpretation.
There’s also a degree of hype around this subject. Exciting as it may be to put on an Oculus Rift at a conference (or more interesting than the rest of the conference anyway) tools like augmented and virtual reality are effectively data and graphics driven, and reflect increased computing power rather than any growth in intelligence. Machine learning is at a more nascent stage, though simple demonstrations make you stop and wonder. Just check out computer scientist Tom Murphy programming a computer to learn how to play Super Mario.
Inevitably, big industries will lead the charge in developing next generation Artificial Intelligence, among them finance, medicine, advertising and FMCG. The technology will be seized upon by start-ups and pornographers who will make more ingenious use of it once it becomes affordable, but the lead will come from those with billions to invest and gain. We’ve already seen a huge and risky transition to the use of trading algorithms in finance.
A further determining factor in the pace of change for artificial intelligence and machine learning will be the adoption, or rejection of open data standards by big corporate, academic and governmental bodies. A paradox that is not well understood by most big businesses is that making their non-confidential data publically available will directly drive insight and opportunity back into the business. Open protocols will surface and solve problems and opportunities that companies don’t even realise exist. Good examples of this include traffic data, delivery logistics and non-confidential medical data.
Our client IBM has been a leading player in AI since the inception of computers, with memorable milestones including Deep Blue beating Garry Kasparov at chess and the Watson supercomputer winning ‘Jeopardy’. At field.work Stephen Tenzer of IBM Watson explained practical applications for cognitive computing including the extraordinary work of Watson in the treatment of cancer. In the past few years Watson’s machine learning capability has enabled it to assimilate and analyse the research of 14 treatment centres in the USA. The initial results include six new possible drug combinations for cancer treatment.
This is a perfect example of Artificial Intelligence because Watson wasn’t able to work from scratch. It needed to learn from decades of research by leading doctors, and thousands of years’ worth of research into cancer, DNA and physiology. But with human guidance the machine was able to produce new and life-changing insights at incredible speed.
As Ken Goldberg argued in his recent piece for the Davos World Forum series, the real focus for businesses and governments in the comings years should not be the ‘Singularity’, but the ‘Multiplicity’. What will we be able to do with the machines once they really are smarter than us? This is a fundamentally optimistic and exciting field of study, and a natural fit with the inventiveness and lateral thinking of the creative industries.
In agencies like Havas helia we’ve made great strides in using cognitive computing and machine learning to interpret consumer behaviour and needs, and fine tune messaging and delivery. The next trend in our world will be to turn our thinking and the power of our robot friends to focus on the classical core of marketing: helping our clients not only with messaging but also with the improvement of their products and services.
By Joe Harrod, content strategist director, Havas helia