Deepfake pioneer Hao Li said that digitally manipulated videos could become ‘perfectly real’ in as early as half-a-year to a year’s time. He cited the emergence of apps such as the Chinese-developed ‘Zao’ and growing research focus on the field. ( MIT Technology Review | Twitter )
A deepfake expert has warned that even regular people will soon be able to create digitally altered videos that look “perfectly real.”
Hao Li, a computer science professor at the University of Southern California, recently discussed the future of deepfake technology.
In an interview with CNBC, Li said that most manipulated videos can still be easily spotted even with the naked eye. However, there are some that have actually become very convincing. He said that these videos often require “sufficient effort” to produce.
According to Li’s estimate, “perfectly real” will be easily accessible to the public in about six to 12 months.
What Is Deepfake Technology?
Deepfake is a portmanteau of the words “deep learning” and “fake” and refers to computer programs that combine human image synthesis with artificial intelligence. The technology is often used to create digital representations or manipulated videos that are made to seem real.
With deepfake technology becoming increasingly more sophisticated over the years, some people are starting to become concerned about its possible negative effects. Digitally altered videos could be used to promote disinformation and confusion among the general public, particularly in the context of global politics as noted by CNBC.
For instance, several social media campaigns and smartphone apps have already been used to spread misinformation, all for the purpose of interrupting elections in different parts of the world.
Emergence Of Digitally Manipulated Videos
Li, who had presented a deepfake of Russian president Vladimir Putin at an MIT tech conference last week, said he initially thought perfect digitally manipulated videos would become reality in two to three years.
However, he later sent out an email explaining that it might actually happen in just half a year to a year.
The deepfake pioneer said he was forced to “recalibrate” his timeline because of recent developments in the technology. He cited the growing popularity of a Chinese-developed app known as Zao, as well as the growing interest of researchers on the field.
“In some ways we already know how to do it,” Li mentioned in the email. He added that the emergence of perfect deepfake is “only a matter of training with more data and implementing it.”
Li warned that there will come a point that people won’t be able to tell which ones are deepfakes and which ones are real anymore. He said this is why there’s a need to look at other types of solutions as well.
ⓒ 2018 TECHTIMES.com All rights reserved. Do not reproduce without permission.