Artificial intelligence is making it easier to construct hyper-realistic digital characters, which has led to a lot of discussion about dangerous and misleading deepfake content. Beneficial use of the technology includes reviving Albert Einstein to teach his famous theory of relativity and discussing career changes with your earlier self.

For this reason, MIT Media Lab researchers and other researchers have developed a freely available, easy-to-use character generation pipeline that combines AI models for facial gestures, voice, and motion to produce a wide range of audio and video outputs. As an extra safeguard against malicious use, the pipeline adds a traceable and human-readable watermark to the final production, distinguishing it from authentic video footage.

The researchers hope that educators, students, and healthcare professionals will be encouraged to use this pipeline if it is freely available. According to research published in Nature Machine Intelligence, students, educators, healthcare practitioners, and therapists could benefit from utilizing these characters.

"When AIs and humans begin to share identities, the world will become rather odd. This article excels at thought leadership while also giving a practical guide for avoiding ethical dilemmas with privacy and misrepresentation," says Jeremy Bailenson, founding director of the Stanford Virtual Human Interaction Lab and an author of the study.

Although the public is most familiar with deepfakes, the paper's first author, Pat Pataranutaporn, a PhD student in professor Pattie Maes' Fluid Interfaces research group, sees "its promise as a tool for artistic expression."

In-depth truths and in-depth learning

As a result of generative adversarial networks (GANs), it has become easier to create lifelike visuals, clone voice recordings, and animate faces. The researchers note that self-exploration is just one of the positive applications of AI-generated characters. Experiments demonstrate, for example, that these characters can increase students' enthusiasm for studying and their success on cognitive tasks. As a complement to traditional training, the technology enables instruction to be "tailored to your interests, idols, and context, and may be adjusted over time," Pataranutaporn explains.

AI-generated content that provides exposure therapy to people suffering from social anxiety. In a related use case, People can use the technology to anonymize faces in the video while preserving facial expressions and emotions. This technology may be beneficial for sessions where participants wish to share sensitive personal information such as health and trauma experiences and for whistleblowers and witness accounts.

However, there are some more artistic and entertaining applications. Students used technology to animate the people in a historical Chinese artwork and develop a dating "breakup simulator" in this fall's Experiments in Deepfakes seminar, taught by Maes and research affiliate Roy Shilkrot.The researchers note in their report that many uses of AI-generated characters create legal and ethical challenges that will need to be considered as the technology develops. Is there a way to figure out who can digitally recreate a historical figure? Who is legally responsible if an artificial intelligence clone of a famous person promotes risky behaviour on the internet? How do we ensure we will not prefer fictional characters to actual ones?

To know more, refer to the article

.

https://news.mit.edu/2021/ai-generated-characters-for-good-1216