Could a Large Language Model Be Conscious?

[extherian]

See here, for an analysis of how far ChatGPT is from being an OA-style turingrade AI.

[stevebowers]

Back in 2002, John B described 'semi-concsious intelligence', which resembles sapience and sophonce on a superficial level, but is base on statistical models and broad sampling databases. This is something similar to what we have today.
https://orionsarm.com/eg-article/47843961d02bf
I think that SI systems will have a lot of utility in the future, and some of these can already pass the Turing Test fairly reliably; but they aren't truly sophont so can't be considered people.

[extherian]

Orion's Arm actually describes the first fully turingrade AI as being developed by around 2050 or so. We are already well on track to hit that target in real life.

[stevebowers]

I'm less optimistic; the evolution of human minds has taken about four billion years, so I expect there are a number of unusual aspects to human consciousness that can't be emulated easily.

But I expect that by 2050 we'll have artificial minds that respond like humans in most detectable ways.