Artificial intelligence (AI) has the potential to significantly boost the UK economy, save lives, and drive cars, a panel of academics told members of the House of Lords this week.
Oxford philosopher Nick Bostrom, and computer science professors Dame Wendy Hall and Michael Wooldridge, were invited to the Houses of Parliament on Tuesday by the Select Committee on Artificial Intelligence.
The trio told the Committee, which is chaired by Lord Clement-Jones, that AI is a powerful technology that must be developed safely and ethically if it is to be of benefit to all of humanity. Bostrom warned that AI could be used in autonomous weapons and for mass surveillance.
“I think the genie is out of the bottle,” said Hall, who teaches computer science at the University of Southampton. “There will be lots of positive benefits in terms of health and discovery of knowledge.
“But the downsides are something we need to get a grip of because it’s happening so fast. There are lots of issues about algorithmic accountability and bias in datasets or in algorithms, and just generally how it’s going to impact society in terms of future jobs. There will be job losses but there will also be lots of new jobs.”
AI is a concept has been around for several decades and we are now in the fourth or fifth wave of AI. The difference today is that there are bigger and better datasets available for AI agents to learn from and machines boast considerably more processing power than they once did.
This has led to a significant acceleration in AI research, led by companies like DeepMind in the UK, as well as US tech giants like Facebook, Google, Amazon, and Microsoft.
The academics told the Committee that the UK is incredibly well placed to lead the AI revolution, adding that countries like China and Canada are also contenders.
“The UK is in a really unusual position that I would not have guessed a decade ago,” said Wooldridge, who heads up the computer science department at the University of Oxford. “We are absolutely at the centre of that AI revolution. This city [London] is at the heart of it. So those opportunities are really in our hand as a nation right now.”
While the academics believe AI has great potential, they also explained that there’s a lot of hype around the technology and that it still has a long way to go.
“Right now there is very limited ability [for AI] to do reasoning, common sense, truly understanding concepts and language,” said Bostrom, who is also the author of the book “Superintelligence.” Bostrom added: “The strengths are in pattern recognition.”
AI today is able to learn how to do certain specific tasks, like play Go or identify an object in a photo, but it can’t excel at several different tasks, and it’s not even close to being able to mimic the human brain.
The breakthroughs have all been in “narrow AI”, but “general AI,” which is the big dream you see in Hollywood, is still a long way off.
“There hasn’t really been any substantial progress in general AI,” said Wooldridge. “We’re beginning to get there with better ideas about the brain but all the progress on AI over the last decade, which is real and substantial and exciting, has been on narrow AI like recognising faces or other specific tasks.”
When asked about whether government should consider regulating AI, the academics all agreed that it was far too early and warned that enforcing laws on AI development at this stage would hinder progress.
Bostrom urged the politicians to look to Canada as a good example of how governments can support AI.
He praised Canada for setting up the $US170 million (£129 million) Vector Institute and launching a $US125 million (£95 million) AI strategy after it noticed several senior academics at top Canadian universities were hired away by US tech giants. The same thing has been happening in the UK, with Oxbridge losing many of its leading experts to DeepMind and other companies like Uber.
“A senior academic in this field could easily move into industry and get that half a million dollar a year salary from many of these big tech companies so to make it attractive for those people to stay in academia one needs something different than business as usual,” said Bostrom.
Wooldridge told the Committee that the UK needs to train up more PhD-level computer scientists and mathematicians, warning that companies like DeepMind are currently dependent on overseas talent.
Hall added: “Because of the situation in the US there’s an opportunity to grasp here. There’s no one running science in the US.”
The AI Committee was appointed on June 29 to consider the economic, ethical, and social implications of advances in AI, and to make recommendations. It has been tasked with producing a report by March 31, 2018.