Deepfake videos "perfectly realistic" are in 6 months: report

The deepfake artist, Hao Li, created this deepfake of Putin, which was shown at an MIT conference this week.AP Photo / Alexander Zemlianichenko; Review of MIT technology

• Deepfake artist Hao Li, who created a Putin deepfake for an MIT conference this week, told CNBC on Friday that "perfectly real" manipulated videos would only be six to twelve months old. .
• Previously, Mr. Li had said that he would expect Deepfakes to be "virtually undetectable" in "a few years".
• When asked for clarification on his schedule, Mr. Li told CNBC that recent developments, including the appearance of the hugely popular Chinese Zao application, had led to "recalibration" its calendar.
• Visit the Business Insider home page for more stories.

In an interview with CNBC on Friday, a deepfake pioneer said that "perfectly real" digitally manipulated videos would only be accessible by an ordinary audience.

"It's still very easy, you can see most deepfakes in the naked eye," said Hao Li, an associate professor of computer science at the University of Southern California, at the conference. CNBC Power Lunch. "But there are also really convincing examples."

He continued, "Soon we will be at the point where we really can not detect [deepfakes] more, then we have to look at other types of solutions ".

Mr. Li created the deepfake of Russian President Vladimir Putin, which was presented at a technical conference organized by MIT this week. The video aimed to show the current state of deepfake technology, which is growing faster than expected. He told MIT Technology Review at that time that the "perfect and virtually undetectable deepfakes" were "in a few years".

When CNBC asked for clarification on his calendar in an email after his interview this week, Mr. Li said that recent developments, including the appearance of the very popular Chinese Zao application, had led to "recalibrate" his schedule.

Read more: The Chinese application Zao viral deepfake allows people to superimpose their faces on celebrities like Leonardo DiCaprio and she is terribly convincing

"In a way, we already know how to do it," he said in an email to CNBC. "[It’s] only a training issue with more data and its implementation ".

Advances in artificial intelligence are increasing the credibility of deepfakes and it is now harder to decipher real videos from falsified videos. This has sounded the alarm about the spread of misinformation, especially in the run-up to the 2020 presidential election.