3,839
edits
Biblical connection - Revelation 13 and Daniel 7: Difference between revisions
Biblical connection - Revelation 13 and Daniel 7 (edit)
Revision as of 12:47, 11 June 2023
, 10 months ago→Revelation 13:5-6 - A mouth is given to the beast: + In March 2023 the first known crime with a digital sound-alike was done as a company was defrauded equivalent of $220,000 with the use of digital sound-alike technology
Juho Kunsola (talk | contribs) (→Revelation 13:5-6 - A mouth is given to the beast: + 2018 Google Research voice-thieving machine required 5 seconds of sample + Microsoft 2023 VALL-E system requires only 3 seconds of sample and can thieve also the "emotional tone" of the sampled speaker) |
Juho Kunsola (talk | contribs) (→Revelation 13:5-6 - A mouth is given to the beast: + In March 2023 the first known crime with a digital sound-alike was done as a company was defrauded equivalent of $220,000 with the use of digital sound-alike technology) |
||
| Line 863: | Line 863: | ||
A little while after the digital look-alikes passed human testing with success humankind created technological solutions with which anybody's naked voice can be stolen from a short sample and to make '''[[digital sound-alikes]]''', which are difficult, if not impossible to distinguish from a genuine recording of that person's voice, and to put those digital sound-alikes to say anything at will. | A little while after the digital look-alikes passed human testing with success humankind created technological solutions with which anybody's naked voice can be stolen from a short sample and to make '''[[digital sound-alikes]]''', which are difficult, if not impossible to distinguish from a genuine recording of that person's voice, and to put those digital sound-alikes to say anything at will. | ||
Earliest known examples of digital sound-alikes are the 2016 [[w:Adobe Voco]] and [[w:Google]] [[w:DeepMind]] [[w:WaveNet]]. These software can also form [[w:phonemes]] which were not present in the training material. | Earliest known examples of digital sound-alikes are the 2016 [[w:Adobe Voco]] and [[w:Google]] [[w:DeepMind]] [[w:WaveNet]]. Adobe Voco required '''20 minutes of sample''' to thieve a voice. These software can also form [[w:phonemes]] which were not present in the training material. | ||
Then in 2018 Google Research researchers presented their digital sound-alike system at the 2018 NeurIPS conference. It required only '''5 seconds of sample''' to thieve someone's voice. | Then in 2018 Google Research researchers presented their digital sound-alike system at the 2018 NeurIPS conference. It required only '''5 seconds of sample''' to thieve someone's voice. | ||
In March 2023 the first known crime with a digital sound-alike was done as a company was defrauded equivalent of $220,000 with the use of digital sound-alike technology. | |||
In January 2023 Microsoft published VALL-E, a voice thieving system that is able to thieve a speaking voice and the ''emotional tone'' of the voice and requires only '''3 seconds of sample'''. | In January 2023 Microsoft published VALL-E, a voice thieving system that is able to thieve a speaking voice and the ''emotional tone'' of the voice and requires only '''3 seconds of sample'''. | ||