Joediaper
Est. Contributor
- Messages
- 696
- Role
- Little
This one brings a whole new meaning to I-Pad...SparkyDog said:this one by far
I don't even have words to explain View attachment 134613
This one brings a whole new meaning to I-Pad...SparkyDog said:this one by far
I don't even have words to explain View attachment 134613
I was going to say, “Or like you’re all left feet?” but that looks like a right foot.SparkyDog said:Ever feel like you're not quite all there? View attachment 134852
You're not wrongLittleAndAlone said:There's no intelligence in AI. It's just trying to match spectral commonalities to things it has been shown and stored in neural net memory in the past. All statistics. It has no idea what it's seeing or doing.
It just sees the ruffles everywhere as high frequency edge data and decides that's a significant recurring feature of "diaper" and weighs it high through repeated neural path reinforcement. So when you ask for a diaper you end up with diapers with 5 leg openings and 3 waistbands.SparkyDog said:You're not wrong
but it's definitely funny to watch the results at times
I think it relies heavily on stock images for its learningLittleAndAlone said:It just sees the ruffles everywhere as high frequency edge data and decides that's a significant recurring feature of "diaper" and weighs it high through repeated neural path reinforcement. So when you ask for a diaper you end up with diapers with 5 leg openings and 3 waistbands.
WTF?SparkyDog said:Ever feel like you're not quite all there? View attachment 134852
You can also over train. Show too many brands or styles or too many aspects or show it pull-ups and disposable diapers and cloth diapers, etc on the same database and accuracy overshoots the bell curve and tanks. Aka diverging vs converging. Need multiple ANNs running in parallel each with their own specialized data subsets for best accuracy.SparkyDog said:I think it relies heavily on stock images for its learning
Yep, the machine learning model is not converging. It’s a problem with AI. Garbage in -> Garbage out.LittleAndAlone said:diverging vs converging
Makes me want to amputate my legs just to try it out!Cottontail said:That last one is genius. I don't know how the diaper companies haven't figured that out yet: If you fill in the leg holes, you can't have leaks there! Duh! (*rushes off to patent this before somebody else does*)
It's good for small narrow focused matching of a specific thing but not broad general use. It's great at profiling infrared or radar silhouettes for target identification when trained with controlled and curated reference data with no noise.Eclectic said:Yep, the machine learning model is not converging. It’s a problem with AI. Garbage in -> Garbage out.
A good effort to thanik veteeras.LilPaddedBunny said:Shockingly bad is an understatement lol. My friend shared this with me a while ago. Glancing at it from a distance doesn't look too bad but the longer you look at it, the more messed up it gets. I like how the kids are popping out of the hood, the guy on the right is leaning out from the wrong direction, and the kid with the sign is decapitated. Better not be late for the "SCHIOOL BUS".
View attachment 134643
At least it did a good job with the Schiool Bus.Riddy said:A good effort to thanik veteeras.
Damn! I think your AI has either been drinking too much or it's tripping on something! LOL!SparkyDog said:This is the closest thing it's come to anything resembling a diaper
looks um uncomfortable?
View attachment 134614
This is why I am not worried about AI replacing anyone’s job anytime soon (or at all).babydl said:Remember that AI is only a computer code that can operate within its limits. Computer code cannot think or realy evaluate most human endeavors. It really shouldn't be given such value and importance.
That's the thing though. It already is, with disastrous (and to an extent, hilarious) results. Companies are putting junk like this out there without thoroughly vetting it... Then they get roasted for it, and start backtracking, but many times not until they've already made adjustments to their workforce.AJFan2020 said:This is why I am not worried about AI replacing anyone’s job anytime soon (or at all).
“I’ll be back”LilPaddedBunny said:Once it becomes self aware you get Skynet and you'd better have programmed a kill switch lol.
We use essential cookies to make this site work, and third party cookies for purposes such as web analytics.