How tech makes faking news easy
As artificial intelligence and real-time video editing improves, telling truth from falsehood is getting tougher than ever
Telling truth from lie is becoming more difficult.
Fake news can already be difficult to discern from truth — but what happens when technology allows for near-perfect, real-time editing of voice and video?
Bias isn’t new in media, nor are incorrect facts, but the past year has seen the idea of fake news come to the forefront, with entirely false news websites set up to skew public opinion or simply to generate ad income.
“The current debate on filter bubbles and social media echo chambers has brought the issue of fake news to much greater prominence,” said Theo Bass, researcher in government innovation at Nesta. “But it has been around for a long time — it has simply been accelerated thanks to technology.”
While it’s easy to fake interviews in text and possible to believably alter still images, doing so in video or audio recordings is more difficult. Advances in technology may change that, with video editing and automated audio mimicry having the potential to wreak havoc on the news industry.
“Adobe’s new ‘Photoshop for audio’ [VoCo] and the Face2Face project’s manipulation of video footage in real-time give us an indication of how the ‘fake news’ environment will become even more challenging to navigate,” said Bass.
Adobe’s VoCo was demonstrated last year, editing recorded speech to construct new phrases to change what was said in a convincing imitation of the original voice. Adobe is already promising watermarked edits to avoid VoCo’s misuse. Face2Face is a research project revealed in a paper last year that can realistically and easily tweak facial gestures in video, letting you put words into someone’s mouth or change their reaction.
Tech, of course, isn’t the only problem. “Equally, we need to address how readily people are willing to accept any news story they read, and how fast this information is able to spread and influence wider opinion,” said Bass. “Though technology is ubiquitous, our ability to either verify or debunk fake news sources will not be the primary issue.”
Fighting back
Bass called on the tech industry to do more to help. “Large companies – whether it’s Facebook or Twitter – often present their platforms as neutral ‘tech platforms’. Yet in reality, they turned into media platforms a long time ago. These businesses need to take steps to be responsible gatekeepers. Clicks cannot be worth more than facts – this will be to the detriment of business credibility in the long run.”
Media companies and activists have stepped up, although such efforts are likely to be limited to people who already trust them. “Slate and The New York Times have created web-browser extensions to alert people to fake news stories, showing the steps media companies can take,” Bass said. “More interesting, however, is that a civil-society group in Taiwan – g0v – has also created a browser plugin, powered by volunteer fact-checkers, which has been downloaded 17,000 times.”
While AI could also be used to help develop fake news, smart algorithms are already being trialled as a possible solution. “In the UK, Full Fact has begun to experiment with automated fact-checking,” he said. “Further afield, the team behind the Fake News Challenge has issued an open call for solutions using machine learning, natural language processing and AI.”
Chris Brauer, director of innovation at Goldsmiths, University of London, pointed out that AI could also help fact-focused journalists do their jobs better. “Media businesses being pressured as they are, there’s been a big reduction in the protecting the validity of sources and being rigorously fact-checked [to] ensure the information is accurate,” he said. “AI will support the journalist in doing the typically difficult work in the trenches – that’s the part the machines are really good at.”
Machine (and human) learning
But if humans can’t distinguish truth from lie, it will be harder for machines to do so. “Success is still a way off — human beings have a hard time distinguishing between real and fake, and AI methods as they are today are not yet sophisticated enough to do this task with a high degree of accuracy.” As Brauer said: “It’s very difficult for machines to filter for the truth, as it were.”
A Nesta research report suggested one answer to fake news online: better digital literacy. “Technology will only ever be part of the solution to what is essentially a social problem,” said Bass. “A push for greater digital literacy and education will be key.”