There are two interesting aspects of #chatgpt that seem to be conflated in many discussions. (Guilty). One is a clever party trick, one may be something more.
One is the GPT-3 part. “credible sounding blather” is up there with “spicy autocomplete” as the best definitions of GPT I’ve seen. It’s a programmer party trick. It’s undoubtedly useful, and will surely be integrated everywhere, like autotranslate quickly was. GPT is no more amazing than autotranslate but generates a thousand times the fuss and hype.
The other aspect is the Chat part of ChatGPT. I understand a lot less about this half, but I gather there are “a hundred” algorithms at work to make it sound like it is conscious. Again this is a party trick, a simulation not a reality. But the actual principle of the Turing Test is that if you can’t tell the difference between a fake consciousness and a real one, then is there really a difference? I’ve seen a couple of discussions about ChatGPT lately that make even me wonder. (Eg this one)