Lancs Uni lecturer weighs in on AI-created fake faces
‘We should be concerned because these synthetic faces are incredibly effective for nefarious purposes’
Research from a Lancaster University lecturer gives new insight into fake faces created by artificial intelligence.
As reported in New Scientist, research conducted by Dr Sophie Nightingale, a psychology lecturer at Lancaster, has found that human faces created using artificial intelligence can look so realistic that people struggle to differentiate them from real faces.
In addition to this, the research also found that people trust fake faces more than real faces.
Dr Nightingale and her colleague Professor Hany Farid at the Univeristy of California, Berkley, asked participants in their study to distinguish 4oo photographs of real people from 400 fake faces. The accuracy rate was found to be 48.2 per cent.
A second group of participants received training in recognising computer-generated faces. Their accuracy was also fairly low at 59 per cent.
Another group of participants were asked to look at a selection of the same faces and rate their level of trustworthiness. On average, the fake faces were rated eight per cent more trustworthy than the real faces. Although this is a small difference, it is statistically significant. One potential explanation for this difference is that fake faces look more like average human faces than real faces do, and people are more likely to trust average-looking faces.
Dr Nightingale highlights how important it is that research is conducted in this area: “We should be concerned because these synthetic faces are incredibly effective for nefarious purposes, for things like revenge porn or fraud, for example”.