So we just need to trick AI into storing them within the model so we can then prompt it to give it back when needed. Seems to work for books and other things they stole...
I opened Facebook. Pit of rage bait mixed with articles designed to look like someone's asking what something is when really it's just engagement bait.
I don't think they're capable of learning from their mistakes, the probability engine and underlying model data will remain the same. It might 'learn' briefly while the conversation history remains current, but will forget it all.
I could be wrong. I just recall trying to teach it something a year or two ago and it was unable to learn.
Some models had a real time network utilization display for the 10mbps bridge and the 100mbps Ethernet. Was great back in the ADSL1 days, could see real time network usage on two sets of four green LEDs.
(And a flashing orange 'collision' light just for fun.)
Soon to be adopted by BMW for their cooling systems?