Posted on: Tuesday 16th of January 2018
AI is a prediction technology. It is a tool we use to mitigate uncertainty about the past, present or future – to understand why something has happened, to coalesce fuzzy data to understand what is happening now and to forecast what is most likely to happen in the future.
AI has traditionally augmented us. Computers handle complex tasks better than we can but stumble on simple tasks that we find second nature – a dichotomy known as Moravec’s Paradox – meaning there should be a comfortable synergy between us.
On the surface, that appears to be changing. Computers’ advantage with complexity remains. AI solutions can track thousands of markers in a slide image to predict whether a patient has cancer or not, but the best pathologists can only track a couple of hundred. Similarly, AI solutions can predict disease susceptibility from genetic code, the likelihood of a specific payment being fraudulent and what songs we will like with greater accuracy than humans.
But the advent of natural language processing and computer vision mean computers are now predicting what a sentence means, what the best reply is, whether there is a cat in the picture – tasks we find easy. And thereby spawning the fear that computers will eventually take over everything we can do – chat bots replacing the UK’s 1 million call centre workers, autonomous vehicles displacing taxi and truck drivers.
Widespread use of self-driving cars won’t happen for another ten years due to the huge volume of solution training required. Similarly, the interactions that chatbots best support – simple information requests, account queries, updates – can be resolved using by existing digital channels. When you consider the profile of who will happily interact with chatbots, it is more digital cannibalising digital than AI replacing humans. Tasks can be automated – not roles – and the multi-tasking nature of most roles is never fully captured in job descriptions.
AI has advanced greatly in areas such as speech transcription and speech generation, lip reading, language translation, object recognition, agricultural yield prediction, crop disease detection, legal research, disease diagnosis and treatment selection. As a result, some specialist roles are being displaced – there may be no great future in becoming a translator. But for the most part, these activities are part of a role and certainly only part of an overall solution. Even disease diagnosis and treatment selection is only part of looking after someone who is sick. And you won’t get a great bedside manner from a computer.
Which brings us to the most vital capability that computers will struggle with – being human. AI solutions can mimic empathy through recognising emotions and altering the interaction respectively. But like a computer-generated painting in the style of Van Gogh, the human element is missing – even only at the perceptual level. If computers are your customers, then AI solutions will serve them well. But for as long as humans are your customers, there will be a part of the relationship that AI will struggle to deliver.