“The people will rise up before the machines do.” According to Andrew McAfee, principal research scientist at MIT, social revolution is more likely long before a Terminator-style takeover by artificial intelligence and robots. Given the relative placidity of the world’s population in the face of evident inequality and greed, that is either reassuring or worrying dependng on your point of view.
The quote was shared by Dr Stephen Cave, executive director at the Leverhulme Centre for the Future of Intelligence, with the intention of arguing that humans will not be enslaved by AI. Speaking in February at The Telegraph Digital Leaders Horizons event, in association with Accenture Digital, he was responding to the accelerated pace of development now being seen. “There has been interest in AI for 60 years, but it now feels different with real and rapid progress. We are seeing the impact of deep learning and data as potential bases for business,” said Cave. “So what does the intelligent future hold for us? Why do we think we need AI to create it?”
Given population growth, he pointed out that there is no shortage of human capacity for tasks. The problem is that people are flawed and there are limits to the capabilities of human intelligence since it inhabits individuals. When a genius dies, so does their genius, Cave pointed out. “To make up for those limitations, AI can enhance those natural processes, just like any other tool,” he said.
What is up for discussion is what type of AI we want and what tasks it is needed to perform. According to Cave, “we want AI that is better than us or there is no point. But that idea makes us anxious - and that fear is not irrational. Intelligence is part of how we identify as a species. So there is a deep ambivalence which we need to understand to make the best out of technology.”
Cave set out two philosophical issues which humans need to confront: the AI superiority paradox and the AI worker paradox.The first of these is the idea of a “godlike machine” which is better at tasks than humans. As the most intelligent species on the planet, this triggers fear about creating something which is superior to us, which we might come to depend on without understanding or controlling.
The second paradox is that, “we want AI to make our lives easier by taking over the drudge work, but then fear being made redundant as a result. The potential is for a social and political crisis if millions of people are put out of work in a short period of time.” That is the risk which McAfee refers to and which is more likely to preceed any enslavement of humans by machines.
The goal of the Leverhulme Centre is to guide AI towards an enhancement, rather than replacement of humanity. That means creating systems which can be trusted without giving away too much because even computers make mistakes, just different ones. “We need to understand the parameters and social dynamics compared to the level of trust. I believe we can create an intelligent future which addresses those technology and social challenges,” said Cave.
One approach being talked about among technology vendors is the concept of augmented intelligence in which machine intelligence supports human decision making. “In the past, humans had to learn the technology and adapt. AI changes that by adapting to us,” said Ray Eitel-Porter, managing director,UKI Analytics, Accenture Digital.
For business, the promise is expanding on the 20 to 30% of value which they are currently unlocking from big data by delving into “dark data” - the huge reservoirs of data being created which are not currently examined or exploited. “AI doesn’t know those boundaries,” he said. If productivity in the world’s top 12 economies could be increased by 40%, then as much growth as has occured in the last 40 years could be gained in the next 20, with half of that uplift the result of augmented intelligence.
Visible evidence of this progress, according to Accenture Digital forecasts, will be that half of all customer interactions with organisations will be with AI bots (without those customers realising it). More challenging is the prediction that voice and gesture will replace screens as the primary interface with computers within seven years. Critically, Eitel-Porter noted, “AI will become your brand.”
As an illustration of the current pace of development, Antoine Blondeau, co-founder and CEO of Sentient Technologies, pointed out that, “eight or nine years ago, nobody was talking about big data and AI.” During that time, his company has raised more funding than any other AI start-up - €143 million - and invested it into building the massive compute power on which AI relies - 5,000 CPUs running simultaneously to support neural network deep learning.
“We think AI has to be part of the decision-making process,” said Blondeau. His company’s approach is configured around lessons from the Korean War that still form part of the US Army’s approach to training soldiers for battlefield decisions based on four steps - observe, orient, decide, act. “The best soldiers are not the most skilled at any one of those - they go through that OODA loop as quickly and as often as possible,” he said. AI is capable of doing that at speeds and cycles far beyond any human intelligence, enabling it to learn even skills which a human has never taught it.
When it came to the live use cases, however, there was the inevitable disappointment that the first place this AI is being deployed is merely digital marketing, in this case optimising the landing page of an ecommrce website where AI was able to drive a 38% higher conversion rate by testing 380,000 different design changes.
When challenged by DataIQ that this seemed a limited application and outcome given the hype around AI, Blondeau responded that the example saw AI supporting human decisions - greater impact would have been seen if AI was given full autonomy. A more interesting example was the AI-guided automated growing of basil at MIT’s agricultural research lab. Allowed a free hand in deciding how much water and nutrient the plants should be given, according to Blondeau, “AI improved the taste of basil.”
The scale of the investment and human intelligence being deployed to make AI a reality is undoubtedly impressive, even if the promised benefits are currently slow to arrive. One reassurance at least emerged from the event - whatever other impacts AI has on the human condition, the future seems set to taste better.