Against the background of two senior cabinet members resigning over Theresa May's proposed Brexit deal, we gathered together four bright minds to discuss what is likely to become a defining argument for the next 20 years: who should have the final say, humans or algorithms?
It seemed only right to ask whether Brexit should have been left to a greater intelligence than ours. "We do a lot of work on stochastic modelling," said Pete Trainor, co-founder of Us Ai and author ofHippo: The Human-Focused Digital Book.
"Think of stochastic modelling as predictive analytics for chaos. We do a lot of work for banks, who need to deal with chaos, and [based on that work] I think if you fed enough data into an algorithm, it would have shown that this was going to be a terrible idea."
The problem said Chloe Grutchfield, co-founder of RedBud Partners, boils down to bias: "An AI doesn't have preconceived ideas. All its decisions are based on a lot of data - and made super-quickly."
This raises the spectre of who supplies the data, but applied futurist Tom Cheesewright was quick to point out that this is a temporary problem.
"Give it a few years, and [AI] goes and gets data. We've already got things called AI systems which don't require programming in terms of their workflow.
"They learn answers from a huge variety of data fed to them, and they work out what's relevant, they can do some level of validation of what's right and what's not by comparing it to other sources. We're not far off basically just throwing the news at an AI, and access to a web search, and asking it to give answers to questions."
This is no distant future, either. "There are already self-trained AI," added Chloe Grutchfield, citing the example of a Polish cyber security startup that analyses workers to create behaviour models. "They look at the operators on the ground that operate satellites they look at the patterns, how do they move their mouse, and how do they operate the machinery, how quickly does the mouse move?
"The AI, the algorithms, can actually identify when a behaviour is not [the operator's] usual behaviour. There is no input of training data, physically, it's just the algorithm learning what the usual pattern of an operator is. And as soon as this operator does something quite odd, he's not moving his mouse as quickly or he's not moving his mouse as much, it says I smell a rat, there is something going on'."
"The secret service across the world has been using this for years," added journalist Sara Driscoll. "They know if an agent is going to go rogue before they go rogue.
"They do analysis of their email, their phone, their patterns. Everyone has a pattern, a way that they work through life, the things they do. Even if they try not to have a pattern, that in itself is a pattern. So AI is self-learning about individuals, and therefore it can indicate if somebody is moving up the going to go rogue' scale."
"It's called cognitive biometrics and most of your banking apps have it in already," said Pete Trainor. "So you put your fingerprint on [your phone] and they identify you, but actually the secondary layer is cognitive biometrics. And if you're holding your phone in a slightly different way it's likely not to be you."
So where does this leave us with Brexit? Let's be contentious for a second. If there is a second referendum, how about we provide an AI-supported decision-making system? One where you say, "I'm worried about immigration", and it sets out the positives and negatives in clear, unemotional terms. And not just for the country as a whole, but right down to your local area.
It's got to be better than relying on mainstream media or our Facebook bubble to help make our decisions - and as it's abundantly clear that our phones know exactly who we are, why not cast the final vote via an app with cognitive biometrics built in?
This article was inspired by just a small section of the discussion from the excellent panellists who joined us at the Mindshare Huddle 2018 last month. They were:
- Sara Driscoll, journalist and consultant
- Chloe Grutchfield, co-founder of RedBud Partners
- Pete Trainor, co-founder of Us Ai
- Tom Cheesewright, applied futurologist
