19-08-2023, 05:23 AM
(18-08-2023, 11:16 AM)Colin Morgan Wrote: Those were the AI's words, not mine. As I understand it, these AI chat things work like lightning fast predictive text algorithms - trying to be helpful but without a clue what they are actually doing.
(It seems that AI chat programs don't KNOW anything about what they are talking about - they work by having access to everything that has ever been written (!) - and then rapidly working out which word would best follow the last, according to an algorithm - a structured set of rules that have been honed to generally give what the programmers think will be the best answer in a given context. AIs quickly but blindly produce text that simulates very well what someone who DOES know what they are talking about might have written. However, they can contain some real howlers and need to be treated with great caution.)
And apparently recently a genealogist asked some genealogy AI for a family history on an obscure relative- the program quickly came back with a family history which was totally incorrect- when checked the AI reply was it could find no record of the person so made the information up to satisfy the question !!
All good to replace proper investigation and evaluation with simplicity
