Artificial intelligence can now make photos talk, we have reached a very dangerous point

Microsoft Research Asia developed by
VASA-1 A new artificial intelligence tool called was introduced. By taking a photo or drawing of a person and using an existing audio file, the tool can create a realistic speaking face in real time. The tool can generate facial expressions and head movements for an existing photo and synchronize lip movements to match a speech or song. Researchers have published many examples on the project’s page, and the results are so successful that they can make people believe that they are real.

However, the lip and head movements in the examples may appear somewhat robotic and out of sync when examined closely. However, it is clear that the technology can be misused; especially real people deepfake
It can be used to create videos easily and quickly. The researchers recognize this potential and have decided not to publish “an online demo, API, product, additional implementation details or any related offerings” until they are confident their technology will be “used responsibly and in compliance with appropriate regulations.” They did not specify whether they would implement specific security measures to prevent malicious actors from using this technology for malicious purposes, such as creating deepfake pornography or misleading information campaigns.

Artificial intelligence can now make photos talk, we have reached a very dangerous point

Researchers believe the technology has many benefits despite its potential for abuse. Technology has the potential to increase equity in education and help people with communication difficulties, even providing them with a tool that can speak for them. avatar They state that it can improve accessibility by providing They also offer the opportunity to provide companionship and therapeutic support to those in need. It is implied that VASA-1 could be used in programs that offer access to artificial intelligence characters that can talk to humans.

Announcement According to the article published with VASA-1, YouTube
It was trained on the VoxCeleb2 Dataset, which contains “more than 1 million utterances for 6,112 celebrities” extracted from their videos. Although the tool is trained on real faces, it also works on artistic photos; researchers demonstrated this in a fun way with a photo of the Mona Lisa combined with Anne Hathaway’s rendition of the viral Lil Wayne song “Paparazzi.” This is worth watching even for those who are skeptical of the positive aspects of the technology.