Artificial Intelligences make deepfakes so perfect even other AIs can't detect them
Artificial Intelligences make deepfakes so perfect even other AIs can't detect them
Deepfake propaganda is
already being used to try to affect the outcome of the US election – but
tomorrow the technology could be used to empty your bank account
Artificial Intelligences are
now so good at creating ultra-realistic deepfake videos that not even
another AI can
detect the deception.
Researcher Nina Schick, whose
book Deep Fakes and the Infocalypse: What You Urgently Need To Know is an
urgent wake-up call about the danger to democracy posed by AI, tells Daily Star
Online: “Humans will never be able to detect deepfakes… it’s already there – to
the naked eye they’re perfect.”
And even AI tools won’t be
able to spot the fakes. She warns: “You might get to the post where the
generators are so perfect that even an AI won’t be able to tell the difference
between a real video and a fake video.
"That’s already the case
with text written by artificial intelligences. No-one can tell…”
AI
deepfakes creating fake humans who 'don't exist' to spread misinformation
A "Swiss security
analyst" named Martin Aspen was behind the alleged leak of secret data
from Hunter Biden – son of US vice-president Joe Biden. But "Martin
Aspen" and his company don't exist.
The photos of Aspen were – as
revealed NBC News – created in a computer. With the US election
on a knife edge, something like the Hunter Biden "scandal" could
quite possibly tip the balance – if it wasn't just based on some photos created
by an AI.
The threat is immense. On a
personal level, people stand to lose their life savings because of rogue AIs
and on a global level political careers can be made or lost, and wars could
start because of convincing faked video footage.
AI deepfake videos to make up '90% of online content' in just five years
Researchers like Nina might
yet provide some defence against the coming information apocalypse. But
equally, she herself admits, they might not.
Luckily, as far as we know
the only people making content with state-of-the art deepfake technology right
now are using it for fun. Sassy Justice is a new viral video show from the
creators of South Park.
The show is “hosted” by a
reporter named Fred Sassy, who appears to be a dead ringer for US president
Donald Trump.
The fakery is so convincing
that it’s easy to forget that there’s any technology involved and somehow the
programme makers have found an actor who’s a long lost relative of the Trumps.
Matt Stone told the New York
Times that the point of the show is to demystify the emerging technology and
make it less frightening. He said: “Before the big scary thing of coronavirus
showed up, everyone was so afraid of deepfakes.“
"We just wanted to make
fun of it because it makes it less scary.”
All the cutting edge technology
didn’t come cheap. Co-creator Trey Parker calls Sassy Justice “probably the
single most expensive YouTube video ever made.”
But the danger is that the
technology is getting cheaper, and quickly. Nvidia’s Maxine uses Deepfake tech
to make video calls look more “natural” and new startup Pinscreen creates
entire digital avatars so that the user can take part in video chats, as Nina
puts it, “without bothering to do their hair or whatever”.
But there’s a lot more to
deepfakes than innocent, useful aspects. In 2016 a gang of conmen stole over
€50 million by posing as French Defence Minister Jean-Yves Le Drian.
They wore simple rubber masks
as they made video calls to wealthy individuals and asked them to fund “secret”
French government missions. They depended on the comparatively lo-fi resolution
of video calls to hid ether fact that they were wearing disguises.
Today, deepfakes would allow
criminals to make photorealistic video calls that even an artificial
intelligence couldn’t tell from the real thing.
It’s already happening. In
March 2019, cybersecurity firm Symantec reported that three major companies had
fallen victim to deepfake fraud, with AI being used to clone voices and call
senior financial officers requesting urgent money transfers. While Symantec
didn’t reveal the names of the businesses, they confirmed that millions of
dollars had been stolen.
The greatest danger, warns
Nina, may not be to huge companies with billions of dollars to lose – and
therefore millions to spend on deep fake detection.
“You know scammers are going
to use it,” she says, “and as it becomes more accessible it won’t be CEOS of
being energy companies being defrauded out of millions of Euros, it’ll be just
ordinary people like you and me.”
Comments
Post a Comment