This year’s Nvidia keynote speech was computer generated, instead of delivered by a human. The voice was provided by an artificial intelligence program called Tacotron 2, which was trained on thousands of hours of speech data.

The results were impressive. The computer-generated voice sounded remarkably natural, and was able to correctly pronounce difficult words and names. It was also able to change its tone of voice, to sound excited, sad, or angry, as needed.

Nvidia’s chief scientist, Jim Houston, said that the computer-generated keynote speech was “the first time that a machine has delivered a keynote at a major technology …