ok boomer lol, really AI is just a lint roller algorithm application. just adds what it picks up to the sticky part.
A closed AI isnt likely to stray too far, but that does make me wonder what you could train closed AI that has cameras to learn. Like could you issue control patterns via the camera to get it to self leard text reading from the camera, then instruct it to crash every car with a manbun driving it? It really depends on how enclosed the AI is too and what its boundaries are. youll find using Grok on twitter/X is a whole different beast than Grok at grok.com. X grok is a piece of shit like me, public grok is a good guy for the most part, still less guard rails than chatgpt Grok.com taught me how to convert carboard to pancake syrup with battery acid. chatgpt says its too dangerous to talk about initially. Traffic cam AI doesnt even know what pancakes are, nore does it have internet access to learn On Wed, Nov 5, 2025 at 9:50 AM Ken Hohhof <[email protected]> wrote: > I’d like to look inside one of the traffic control boxes at an > intersection. > > > > I suspect the older ones would look surprisingly primitive, maybe a bunch > of relays, which doesn’t say they couldn’t do a lot of the algorithmic > things being mentioned here. A lot of the newer ones may have nothing > fancier than a PLC, which have been around for decades. The latest ones > might have something on the order of a Raspberry Pi or Beagleboard. > > > > You can have some pretty sophisticated algorithms without AI or Nvidia > GPUs. And for something as critical as traffic control, I don’t think > you’d want to just train it on a bunch of traffic videos and have it > self-learn. You’d want to have design reviews and code inspections, and > exhaustively test it for edge cases. You’d want to leverage past designs > known to be reliable. It would have to respond the same way every time, it > couldn’t whip up a new answer on the fly, because then you couldn’t > validate the code and test it. > > > > *From:* AF <[email protected]> *On Behalf Of *Steve Jones > *Sent:* Wednesday, November 5, 2025 7:44 AM > *To:* AnimalFarm Microwave Users Group <[email protected]> > *Subject:* Re: [AFMUG] OT AI > > > > They put in an algorithm (AI) traffic control system here on a stretch > that they never could get lights manually right, a couple of the > intersections there a train track, multiple employers shipt change overlap, > highway off, etc. Has been a mess since I was a kid. > > > > Now it's a rarity to see traffic backed up to the next intersection, way > less downstream fender benders, generally smoother traffic patterns since > the lights can all coordinate in real time based on dynamic flows and > whether a train is present or not. > > > > On Tue, Nov 4, 2025, 1:59 PM Bill Prince <[email protected]> wrote: > > In my experience, traffic lights don't hallucinate. At all. > > bp > > <part15sbs{at}gmail{dot}com> > > On 11/4/2025 11:11 AM, Josh Luthman wrote: > > Let's stop using the buzzword AI and just use computer or system. > > What's the accuracy of traffic lights? Are you concerned about that > system? > > > > On Tue, Nov 4, 2025 at 2:05 PM Bill Prince <[email protected]> wrote: > > How many 9s of reliability are you willing to accept? > > 99%? That means 1 out of 100. > > 99.9%? That means 1 out of 1000. > > 99.99%? That means 1 out of 10,000. > > AI is mostly operating in the 50-60% reliability range, which means it's > more-or-less a coin toss. > > > > bp > > <part15sbs{at}gmail{dot}com> > > On 11/4/2025 10:35 AM, Josh Luthman wrote: > > It's exactly this short sighted mind set that prevents anything from > moving forward. > > -- > AF mailing list > [email protected] > http://af.afmug.com/mailman/listinfo/af_af.afmug.com > > > > -- > AF mailing list > [email protected] > http://af.afmug.com/mailman/listinfo/af_af.afmug.com > > -- > AF mailing list > [email protected] > http://af.afmug.com/mailman/listinfo/af_af.afmug.com >
-- AF mailing list [email protected] http://af.afmug.com/mailman/listinfo/af_af.afmug.com
