A Streaming AI Musical Christmas Card From the Future

[Ace-Garageguy]

 

[Monty]

Every bit as enjoyable as the last one you posted.  I was curious about something though.  Why do you think there were so many intentionally misspelled words?  For example, look at the wording on the front of Santa's rocket sled.  It's not like they're trying to avoid copyright issues 'cuz those wouldn't come into play here.  

[Ace-Garageguy]

2 hours ago, Monty said:

... Why do you think there were so many intentionally misspelled words?  For example, look at the wording on the front of Santa's rocket sled.  It's not like they're trying to avoid copyright issues 'cuz those wouldn't come into play here.  

I would guess the AI doesn't recognize them as words specifically, but "thinks" they're just designs it's seen something similar to, and doesn't put them in context as images that should be interpreted as text, so as to make the reasoning leap that proper spelling is part of the equation.

The reason I think this is because there are other words that look like words at first glance, but if you look closer, or stop the frame, they're just squiggles, or partial words with some squiggles.

Some people have difficulty with interpreting stylized typefaces as words occasionally, so it's not surprising a machine might also.

EDIT: There some that don't fit that supposition well though, like the "guaramted midmight delivery" or something similar on a Santa frame. Poor spelling ability on the part of the human(s) behind the production? Not a first-language-English speaker? Sloppy editing? Somebody just being oh-so-cutesy? Who knows?

Still, from what I'm seeing, these consumer- and art-grade AI products are not really good yet at understanding context and nuance, or making logical assumptions that human minds do effortlessly (some human minds...). These abilities can, however, be defined as more sophisticated pattern-recognition, and will continue to develop over time.

This one seems to me to be just a tick better than the last one. The last one had all kinds of flaws, like vehicles moving kinda sideways, and licking ice cream or eating being rendered very oddly. Maybe it's learning.  

EDIT: It's a failure of AI to grasp context and make logical assumptions that (some) humans generally excel at subconsciously that seem to be responsible for the mistakes self-driving systems make as well.

Edited December 8, 2024 by Ace-Garageguy

[espo]

60 Minutes had a program about the companies doing the programing for AI in Africa. The main theme had to do with the company being hired to git programmers and was paying them around $12 or $13 per hour. That company was hiring locals in Africa and was only paying them pennies. 60 Minutes interviewed several of the programmers and got information on how and what they were doing.   I think you can find the program on the internet somewhere if that would be of interest to anyone.