Technology
Experiments reveal that audiences can generally determine video deepfakes of popular political leaders– however phony audio and text are more difficult to identify
By Jeremy Hsu
Facebook/ Meta Twitter/ X icon Linkedin Reddit Email
This is a genuine picture of Joe Biden providing a speech
SHAWN THEW/EPA-EFE/Shutterstock
Individuals can typically identify when videos of popular political leaders offering speeches are really AI-generated deepfakes. We have more difficulty critical fakes from truth when listening to audio or reading expected text records.
“Audio deepfakes are, in my viewpoint, a little bit more hazardous in the present time due to the fact that visual deepfakes are still more difficult to produce,” states Aruna Sankaranarayanan at the Massachusetts Institute of Technology.
Sankaranarayanan and her coworkers gathered text records, audio and video of political speeches …
Open this short article
No dedication, cancel anytime *
Deal ends 15 December 2024.
* Cancel anytime within 14 days of payment to get a refund on unserved problems.
Inclusive of appropriate taxes (VAT)
or
Existing customers
Check in to your account
More from New Scientist
Check out the most recent news, posts and functions