The REALLY impressive thing to me (but there is more than one) is that it absolutely' knows' what you want when you ask a detailed question. I put 'knows' in quotes because it doesn't know like you and me, but it gets the point almost perfectly, unless you ask the question in too brief a way - but then that would happen to a human too. It is completely tuned to 'getting the point' about what you want. This is a VERY sophisticated thing. Plenty of people are bad at that. I've met loads of them. As a tool of usefulness, producing a completely relevant answer is a fantastic asset for answering queries.
Another impressive achievement is the way in which it operates 'like a human'. It is able to present human-like manners and references. It gets the joke if you tell it one and it gets irony too - another thing plenty of humans can't do. I don't know if you read the link, but at the end of its answer to by two part question, I complimented it on doing a good job, and its reaction was extremely human like.
Then impressed as I was, I wondered what Alan Turing - sometimes called the father of computing, would have thought of its present abilities, and it went off into an interesting musing about that - with full and detailed knowledge of the issues Turing was interested in.
AS a person whose first experience of owning a computer was buying a ZX81 when they first came out in the 1980s and being profoundly disappointed at how useless it was at doing anything helpful, I am amazed at the progress since then - especially at the way these large language models power of usefulness is expanding exponentially. Five years ago, I don't think we could have expected the rate of progress we have seen. The progress in the way the models perform since when Chat-GPT first came out is unexpected - at least to me.
oilprice.com
www.intellinews.com
youtube.com
