interesting thread…one thing hasnt been touched on, ai also has other human features, it can get lazy, it can repeat the same mistakes over and over, and although the good news is it isnt really repeatable, the bad news is it isnt repeatable. and there is always the possibility of falling into an error “hole” that wasnt accounted for in the data set. so yes its scary but it also has some big limitations that require a lot of oversight. for me its the greatest thing since sliced bread. it allows an individual to compete with the big boys which means that they will figure out a way to get all the sheep to sign up to limit its usage by the small guys and let the big boys keep playing. same stuff different day…lol
Indeed it will repeat the same mistakes, unless it’s trained otherwise…
Bayesian Spam filters were a pretty good early example. If you regularly checked the junk folder to correct anomalies they quickly learnt your preferences & did a good job. If you didn’t do any checking then naturally they kept making mistakes; and every uncorrected ‘same mistake’ made that mistake seem more ‘correct’ to the filter algorithm and thus more likely to be repeated…
i’m not doing my own networks…lol