‘Fake news not killer robots is real AI threat’
Expert calls for EU-wide vigilance over ‘existential’ danger to elections
It’s Leo VaradkAI : We asked an AI art generator to illustrate this story with An Taoiseach as a killer robot with an ‘AI eye’ – and this is what it produced
THE potential for artificial intelligence to spread disinformation during election campaigns is a far greater threat to society than the ‘risk of killer robots coming up the garden path’, according to Ireland’s foremost AI expert.
The warning comes as we face into a year of local and European elections, and a possible general election, as well as a presidential vote in the US and a probable national poll in the UK.
Professor Barry O’Sullivan, who last year became the first Irish person to win the European Association for Artificial Intelligence’s EurAI Distinguished
Service Award, has played down the ‘killer robot’ risk of artificial intelligence, which he says gets too much coverage.
Instead, the University College Cork professor said it is AI’s ‘impact on democracy and the wellbeing of society’ that are ‘the things that we need to worry about now’.
Professor O’Sullivan told the Irish Mail on Sunday: ‘If we get elections wrong and you put lunatics in power, then you could find yourself in war situations and that becomes existential.’
AI has been employed during previous election campaigns, but rapid advances in generative AI technology in the past couple of years have left experts fearing that 2024 will usher in a whole new era of disinformation – one in which people are unable to believe their own eyes and ears.
AI software can now create an image or advanced piece of writing out of nothing but simple instructions.
The technology is convincing enough that the Irish Times was tricked into publishing an article last year supposedly written by a freelance journalist, when in fact both the article and the author’s photo turned out to have been created using AI. And ‘deepfake’ technology means computers can learn from existing images, video and audio and then mock up new versions so convincing they are almost impossible for the untrained eye or ear to spot. Prof O’Sullivan gave the example of images that circulated last year appearing to show Pope Francis wearing a puffer jacket.
‘They were fairly harmless in some sense, but it was interesting in that people were debating whether the photos were real or not. The problem is, you might not know if it’s real or not unless you’re actually told.’ Taoiseach Leo Varadkar fell foul of deepfake technology last year when a pair of Russian comedians posing as an African diplomat tricked him into participating in an online call, and then manipulated the resulting footage. A spokesman said at the time that the call ‘transpired to be a sophisticated deepfake’.
AI has been used in previous elections in ways that seem relatively benign compared to the ways it can mislead today.
It can be and has been used to identify undecided voters and create bespoke advertising campaigns. And it has been used to write speeches that will appeal to a particular audience.
On the more nefarious side, AI has been used to mobilise armies of bots – fake social media accounts that post automatically according to certain trigger words written by other accounts, either supporting or rubbishing the original post, depending on the intention of the bot’s creator.
These bots are often used to sow division and can create a skewed impression of what public opinion truly is.
‘That’s one of the sinister ones because it really stifles debate,’ Prof O’Sullivan said, adding that such accounts are often ‘easy to identify’ on platforms such as X (formerly Twitter).
‘The profile photograph is not of a person,’ he advised. ‘There’s lots of numbers in the username. They don’t have a lot of followers. They aren’t following many and often they haven’t existed very long.’
Prof O’Sullivan also suspects there was a lot of non-human influence surrounding the recent riots in Dublin, with bots helping to stir up anger among people. AIgenerated images of Conor McGregor leading an angry mob also went viral in the hours after the riots. While the pictures were clearly not real, they serve to ‘whip up feeling in people’, Prof O’Sullivan said.
‘It’s clearly toxic – this notion of rebellion and Ireland for the Irish. You see this a lot now on social media – people making these statements like “Home the Irish homeless first”, “Ireland for the Irish.”
‘You might see it from your aunt Aggie or someone who clearly isn’t an extremist in any way. But the argument makes sense to her, so she doesn’t think it through. Then she shares it and then some other person shares it.
‘So these kinds of things exploit people who don’t think about the issues very deeply and that then gets permeated into the society. And while it might have started from an extreme group or a very toxic group trying to push a very racist or extreme nationalistic view, it can become mainstreamed by the ordinary person just sharing it and saying, “Well, why isn’t it like this?”’ he said.
In response to queries from the MoS, a spokeswoman for Coimisiún na Meán, the new regulator for broadcasting and online media, said the ‘integrity of elections and the integrity of civic discourse are critically important to society because they are at the heart of what makes us a democracy’.
The spokeswoman added that, when the EU’s Digital Services Act comes into force next month, the commission will ensure platforms are adhering to their own terms and conditions – ‘including those on how they deal with misinformation and disinformation, political advertising and hate speech’ and she said it will ‘take compliance action if they are not’.
‘The question of electoral integrity is a concern across the EU. Many platforms have signed up to a code of practice on disinformation which sets out the ways in which they will counter this threat. We expect platforms to be aware of the risks of AI and AI-generated content and to use systems and processes to deal with it,’ the spokeswoman added.
However, Prof O’Sullivan believes the AI genie is already out of the bottle for elections this year, and said the EU needs to develop a verification system for media similar to the Farm To Fork initiative for food, before things go too far.
But he believes Ireland is less susceptible to malicious online interference than other countries.
‘Despite the fact that we think we’re a very digital economy, I’m not sure we are, really,’ he said. ‘I think the ordinary person on the street probably isn’t reading social media for their news as much as we think they are. And so, there’s probably a huge subset of the Irish population that just isn’t accessible or manipulatable. You could not influence my mother, for example, if you’re a Russian bot, or a Russian tech company trying to influence the Irish election, because she just wouldn’t see your content.’
‘You could find yourself in war situations’
‘AI has been used to mobilise armies of bots’
‘You could not influence my mother, for example’