3,958
edits
Synthetic human-like fakes: Difference between revisions
→Timeline of synthetic human-like fakes: + 2023 | March 7 th | science / demonstration | Microsoft researchers submitted a paper for publication outlining their Cross-lingual neural codec language modeling system at arxiv.org dubbed VALL-E X at microsoft.com
Juho Kunsola (talk | contribs) ('terroristic synthetic pornography' is more correct language-wise than the previous wording) |
Juho Kunsola (talk | contribs) (→Timeline of synthetic human-like fakes: + 2023 | March 7 th | science / demonstration | Microsoft researchers submitted a paper for publication outlining their Cross-lingual neural codec language modeling system at arxiv.org dubbed VALL-E X at microsoft.com) |
||
| Line 458: | Line 458: | ||
== 2020's synthetic human-like fakes == | == 2020's synthetic human-like fakes == | ||
* '''2023''' '''<font color="orange">Real-time digital look-and-sound-alike crime</font>''' In April a man in northern China was defrauded of 4.3 million yuan by a criminal employing a digital look-and-sound-alike pretending to be his friend on a video call.<ref name="Reuters real-time digital look-and-sound-alike crime 2023"/> | * '''2023''' | '''<font color="orange">Real-time digital look-and-sound-alike crime</font>''' | In April a man in northern China was defrauded of 4.3 million yuan by a criminal employing a digital look-and-sound-alike pretending to be his friend on a video call.<ref name="Reuters real-time digital look-and-sound-alike crime 2023"/> | ||
* '''2023''' | '''<font color="orange">Election meddling with digital look-alikes</font>''' | The [[w:2023 Turkish presidential election]] saw numerous deepfake controversies. | * '''2023''' | '''<font color="orange">Election meddling with digital look-alikes</font>''' | The [[w:2023 Turkish presidential election]] saw numerous deepfake controversies. | ||
| Line 464: | Line 464: | ||
** "''Ahead of the election in Turkey, President Recep Tayyip Erdogan showed a video linking his main challenger Kemal Kilicdaroglu to the militant Kurdish organization PKK.''" [...] "''Research by DW's fact-checking team in cooperation with DW's Turkish service shows that the video at the campaign rally was '''manipulated''' by '''combining two separate videos''' with totally different backgrounds and content.''" [https://www.dw.com/en/fact-check-turkeys-erdogan-shows-false-kilicdaroglu-video/a-65554034 reports dw.com] | ** "''Ahead of the election in Turkey, President Recep Tayyip Erdogan showed a video linking his main challenger Kemal Kilicdaroglu to the militant Kurdish organization PKK.''" [...] "''Research by DW's fact-checking team in cooperation with DW's Turkish service shows that the video at the campaign rally was '''manipulated''' by '''combining two separate videos''' with totally different backgrounds and content.''" [https://www.dw.com/en/fact-check-turkeys-erdogan-shows-false-kilicdaroglu-video/a-65554034 reports dw.com] | ||
* '''2023''' | January 5th | '''<font color="red">science / demonstration</font>''' Microsoft researchers announced [https://www.microsoft.com/en-us/research/project/vall-e/ '''''VALL-E''''' - Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers (at microsoft.com)] that is able to thieve a voice from only '''3 seconds of sample''' and it is also able to mimic the "''emotional tone''" of the sample the synthesis if produced of.<ref> | * '''2023''' | March 7 th | '''<font color="red">science / demonstration</font>''' | Microsoft researchers submitted a paper for publication outlining their [https://arxiv.org/abs/2303.03926 '''Cross-lingual neural codec language modeling system''' at arxiv.org] dubbed [https://www.microsoft.com/en-us/research/project/vall-e-x/vall-e-x/ '''VALL-E X''' at microsoft.com], that extends upon VALL-E's capabilities to be cross-lingual and also maintaining the same "''emotional tone''" from sample to fake. | ||
* '''2023''' | January 5th | '''<font color="red">science / demonstration</font>''' | Microsoft researchers announced [https://www.microsoft.com/en-us/research/project/vall-e/ '''''VALL-E''''' - Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers (at microsoft.com)] that is able to thieve a voice from only '''3 seconds of sample''' and it is also able to mimic the "''emotional tone''" of the sample the synthesis if produced of.<ref> | |||
{{cite web | {{cite web | ||
| url = https://arstechnica.com/information-technology/2023/01/microsofts-new-ai-can-simulate-anyones-voice-with-3-seconds-of-audio/ | | url = https://arstechnica.com/information-technology/2023/01/microsofts-new-ai-can-simulate-anyones-voice-with-3-seconds-of-audio/ | ||