Pioneering Scotland can create 'superhero sidekick' AI but challenges remain
FORGET Batman and Wonder Woman – mere mortals are about to gain superpowers through Artificial Intelligence (AI).
That’s the prediction of a scientist at Edinburgh University which this year celebrates its 60th anniversary of being a global leader in AI.
Drew Hemment, Professor in Data Arts and Society, said the arrival of the new technology would give us new capabilities.
“In this new generation of AI it’s like we get an assistant that can help us do things we could never do alone,” he said. “The AI is our superhero sidekick.
“The arrival of AI is as profound a moment as the arrival of the internet or the mobile phone.
“Thanks to developments in the field, ordinary people will soon be able to paint like Peploe, compose poetry like Burns or write novels like Stevenson.
“But before we get too carried away, the AI only generates an approximation, or likeness. The AI needs a human collaborator with ideas and intent to produce really significant art.”
Hemment added that there are already some industries where using these kind of tools is already the norm.
“We really are in for a profound shift so we need to strap in tight and be ready for the ride,” Hemment continued.
One of the concerns around AI which is “partly justified and partly not” is around job losses.
“It is a bit of a misrepresentation because whenever we have had new technologies there have been fears around job losses but what we tend to see is that new technology takes some roles away but creates new ones too,” said Hemment.
The possibilities will be debated by experts in the field this week at the Scottish AI Summit in Glasgow.
Hemment said that at the moment the outlook for AI in Scotland and around the world was a “fascinating mix of positive potential” and “some very troubling aspects”.
Positive breakthroughs are happening now, particularly in gene therapy where there has been huge advances but negative aspects of the new technologies need to be “carefully negotiated”.
The dominant form of AI is machine learning which involves spotting patterns in vast troves of data. Generative AI is a recent development which can generate new things – such as a poem or an essay – by analysing patterns in historic data.
However, because the machines are trained on data harvested from society, they reflect any inequalities embedded in society.
“Society is a deeply troubled place with lots of inequalities so these AI technologies that are trained on the data reflect and amplify the biases and inequalities in society,” Hemment pointed out.
“There is a very famous case of a self-driving car being trained on a data set predominantly of Caucasian people and not recognising Black faces.”
The technology is also increasingly used in areas like credit ratings or by insurance companies, so if the data is biased against race, class or gender that can affect people’s access to services, finance and insurance.
“That is very real and very now,” said Hemment. “It is not a far future challenge.
“Another is in policing – less so in Scotland – but in the US facial recognition technology is being used and if it is biased that feeds into the policing in problematic ways. Regulation is needed and we should not be afraid to say there are some areas where we don’t need or want these technologies.”
Another problem is image generators as they are trained on images taken from the internet generated by real people without their awareness and without any payment.
There are also issues around the way the data is obtained as a lot of it is from extractive business models, as seen in services like Facebook.
“They harvest data about our interactions and use it for reasons that are not transparent to us,”
said Hemment. “So there are a number of troubling issues but there is important work happening now in highlighting and tackling those challenges.
“It is an exciting moment but also very sobering because of these negatives right in front of us that are very real and very immediate. We need to address some of these key foundational challenges in machine learning so that we can have more ecological, more ethical and inclusive technologies in the future.”
The science has come a long way from the 1960s when Donald Michie, a code breaker at Bletchley Park during the Second World War, set up an AI lab in Edinburgh – one of only four in the world and the first in Europe.
“It was the home of AI in Europe and it trained many of the key AI scientists that went on to work at Google so it has a hugely influential history,” said Hemment.
Edinburgh was not only a pioneer in the field of AI but stayed committed to the science, even in periods when fashions changed and funding dried up.
“It was one of the few that kept going through those ‘AI winters’ when it wasn’t fashionable and kept on making those breakthroughs and doing hugely significant work,” he said.
“It is the fact that it is a pioneer but also that it was consistent, stayed committed and stayed at the forefront of all these incredible breakthroughs for decades that really makes it unique.”
As well as building the MENACE computer in the 60s, Michie worked on very early robots that were the most advanced of their kind at the time.
He also set up the world’s first AI spin out company in 1969 as well as achieving other early milestones in the history of computing.
That innovation has continued with Hemment being in charge of the innovative New Real Hub for AI and arts research, while the School of Informatics is world leading and Edinburgh has also recently established labs for integrated AI and quantum software.
Brexit, however, has had a “chilling effect” and has had a major impact both in terms of funding and maintaining links with European colleagues, according to Hemment.
“A lot of my own research was with European labs and that side of things has shrunk,” he said. “We go on and where possible we find ways to continue to collaborate with European colleagues. The belief in internationalism still burns strong but Brexit has made it harder and some of those collaborations with European colleagues without doubt have had a knock so it has had a chilling effect.
“However I always try to be an optimist and hopefully the sands will shift again, whether that is through Scotland in the future or something else. In the meantime I am a believer in nurturing those links and trying harder – you have got to keep battling if you believe in those principles and values.”
The Scottish AI Summit (www.scottishaisummit.com) takes place March 28-29 in Glasgow, where Professor Hemment will present the latest work by The New Real (www.newreal.cc)