(01-05-2026, 05:02 PM)Mark Knowles wrote: You are not allowed to view links. Register or Login to view.It is easy to criticise these AIs, but I think they clearly can be useful tools. Even if 95% of what they say is nonsense that remaining 5% may be gold. I suppose they just need to be used with care and caution like any tool should be.
Identifying the Andrea Barbarigo cipher was a useful lead for me to follow up. I haven't seen it yet, but I doubt Ioanna Iordanou made it up.
I am hoping they will improve but right now I think they can cause a lot of harm insofar as their pandering and especially insofar as incorrect information.
Speaking of harm, I hear there will be a class action suit against Open AI regarding the families who lost people in the Tumbler Ridge shooting, because they had banned the account of the shooter but had not advised law enforcement of their plans, despite many employees thinking they should. Would be interesting to know the conversations there and how the AI responded, ie how involved might those conversations have been in terms of leading to the actual performance of the plans, including the banning. Hoping some protocols will be developed along the way to improve the systems in that way at least.
I heard yesterday that they have been doing tests with AI with regard to medical differential diagnoses and they are doing very well there, agreeing with emergency room doctors 84% of the time, i think they said. I think i would trust that type of ai answer more because of the closed data set involved, and the lack of personalization involved, both in terms of the ai, and in how human doctors sometimes allow human factors in where they don't belong.
I guess that is the answer, as rikforto mentioned, refine the search!