The AI ​​can imitate any sound after 3 seconds

Artificial intelligence (AI) and other software can mimic sounds that have been around for some time. But this usually requires a lot of source material so that the program can learn exactly how someone pronounces each sound. But the new Val-E From Microsoft it only takes three seconds.

After a short, random clip of audio, VALL-E can mimic the voice, pitch, and pronunciation of the voice. AI tested by Cornell Universitywhich some examples Involved.

Some of these examples sound natural and convincing, and others are as automated as older attempts at artificial intelligence.

Not yet available to the public

VALL-E has been trained with 60,000 hours of audio recordings. Like other AIs, VALL-E only gets better at its functionality as more materials become available. VALL-E is not yet available to the public.

This type of AI is controversial because it is easy to use and above all to abuse. Thus a person can make a statement that is indistinguishable from another person

ChatGPT in Word and Powerpoint

Meanwhile, Microsoft plans to release AI text chat to be incorporated into their products. ChatGPT is a chatbot that gives answers in natural language, also in Dutch.

ChatGPT will be included in Office Word, Powerpoint and Outlook products the information. The software could, for example, soon be able to write emails in the user’s tone.

Microsoft wants a chatbot, too for the Bing search engine To give users faster answers to their questions. With the function of artificial intelligence, Microsoft wants to step up the fight against Google.