As soon as the beta version of the neural network from the startup ElevenLabs, presented to users, has already become the object of hacking. Currently, developers are collecting information about abuses and are working to find ways to protect the Voice Lab service from Internet hooligans.
The fact is that a neural network that can voice a text with the maximum similarity to any voice is used to fake celebrity voices. It was their voices that voiced offensive, racist and homophobic slogans that appeared on the infamous resource 4chan, writes Engadget.
The publication cites examples such as a clip with a voice similar to that of Emma Watson, who reads an excerpt from “Mein Kampf.” In addition to the cinematic Hermione, dubious messages were voiced by “Quentin Tarantino” and “George Lucas”.
To protect the Voice Lab service from voice bullies, ElevenLabs plans to introduce additional account verification.Also, the startup will require users to confirm ownership of the cloning voice. To do this, they will have to read the provided sample text.
Finally, the company may opt out of the Voice Lab service. Users will then have to submit voice cloning requests that will be manually checked.
Recall that scientists from the University of Minnesota commissioned artificial intelligence ChatGPT to prepare answers for final exams in four disciplines. According to the results of the study, it turned out that ChatGPT without human help is not able to achieve successful results.