As part of a tour to promote her new book, Dr Hannah Fry took part in a discussion at the Royal Society of Arts last week. Some of the opinions she expressed included; why we don’t have to worry about an android-powered apocalypse for many years, why the way we communicate now is damaging democracy, and why "intelligent assistance" might be a more useful term than "artificial intelligence".
"There are more pressing concerns than evil AI."
A question on the minds of at least two attendees was whether artificial intelligence will lead to the demise of the human race. Fry repeated the analogy of Stanford University professor Andrew Ng who said in 2015 that worrying about evil AI is akin to worrying about overcrowding on Mars. She went on to say: “No-one has even replicated the brain of a worm which has 280 neurons, compared to the billions that we have in our brains. We are a long, long way away, so there are a lot more pressing concerns.”
When another attendee challenged this response by saying that many other AI researchers have criticised the Mars overcrowding analogy and that rather than being evil, there is valid concern about the software having bugs, Fry responded that those expressing those beliefs may have an ulterior motive and that are profits to be made from people believing that evil AI is around the corner. She said: “If you’re selling driverless cars, it is profitable for people to believe that you are the people who really understand how sentient AI works.”
"For democracy, filter bubbles are really, really dangerous."
The “datafication of elections” is an issue we’ve covered before at DataIQ and is concerned with filter bubbles that mean people see extremely personalised political ads that tap into their fears and concerns. In the context of this, one attendee asked Fry to comment on whether AI will destroy democracy. She admitted that she is worried about that.
In her view, in the recent past everyone had the same access to information about politicians and their parties and policies. Nowadays, however, Facebook is one of the biggest portals through which people get their news. “When you have these very personalised newsfeeds that are dictated on what you like or dislike, these incredibly splintered conversations happen. When you have conversations with people, there’s something slightly mismatched and you’re arguing about different topics. For democracy, that’s really, really dangerous because you can’t have sensible, national debate unless everyone has the same information.”
"The words AI add 10% to your funding."
One attendee made the point that the term "artificial intelligence" might be better understood and accepted if it were instead referred to a "decision support system". Another went even further and suggested that it be called "intelligent assistance" and Fry wholeheartedly agreed. She said: “What we’ve seen so far isn’t a revolution in artificial intelligence. It’s a revolution in computational statistics. Unfortunately, that is a lot less sexy.” She added that many people would be less likely to fund projects if they didn’t have AI in the title and that, “the words AI add 10% to the funding you can command.”
Fry’s book is called "Hello world: Being buman in the age of algorithms", and so one talk attendee naturally wondered about similarities with Cathy O’Neil’s Weapons of Math Destruction. She responded: “Cathy’s book is needed. It said, ‘here are horrible, horrible things that lie in our future.’ What I do in my book is take a much more measured approach. I try and be very balanced because there are enormous positives to be had, just as long as we can find a way through this tangled web. I am offering suggestions as to how we might navigate that.”