Quote:
Originally Posted by itshotinvegas
Have you ever used the voice-to-text feature on your keyboard to make one of your long post? Try it, and imagine it trying to figure out, and convey context.
Voice-to-text is a fairly different problem domain from AI text generation, fwiw, although they share in common some basic problems in natural language processing.
But if anything the recent history of voice-to-text would suggest more concern, rather than less, if you want to connect them. Machine learning techniques have improved that tech by an amazing amount. I did a small hackathon project last year using Amazon Transcribe to process audio clips recorded as part of emergency medical care. It was definitely one of those "the last 10% to make this viable as a product will take 90% of the work" kinds of projects, but it's pretty impressive how far you can get in even just a few days.
But yeah, I dunno, deep fakes and AI personas and dystopian sci-fi novel plots. I'd be hesitant to make any sweeping pronouncements. But it's not
that implausible given existing tech.