There's tons of this AI-generated muck on YouTube. It's not involving any kind of IP, or else it would probably get taken down. However, while it looks clean, the movement and speech of the people in it is downright terrible. .
my last post was actually referring to a faux 4k up-scaled version of the original episodes
since i wrote that, i did come across a software solution randomly in a social media feed, but i doubt it would be good enough on some of the crappy quality stuff we have as source. But from a high quality company like topaz or google or microsoft, i think they'd be able to do a remarkable job at some point.
]]>Where the "value" might be is in recording dialogue from actual actors, written by actual humans, and then having AI animate the thing, so that they "look" like the original thing. Again, you'd be sued for trying it, but that's something of interest. But probably still several years away PLUS those models are not cheap to use.
]]>I would welcome it. Though I am not sure the studios would do that, considering AI should theoretically be able to do such a good job, I don't see why they wouldn't for such savvaged properties that have already been compromised from their original image anyway. AI should be able to do a good job because most of the detail is there to begin with.... it's not like they will be inventing new objects other than maybe some background objects as it tries to guess on clarity)
]]>AI was able to handle monitoring and restocking inventory of a vending machine. What it couldn't handle was product selection, individual customer product requests, calculating pricing against competing vendors, weighing one customer request against other different requests. It could vend the drinks, but it was unable to make decisions across multiple variables and overlapping situations.
In terms of medications and solutions, I think AI is as it exists is good for generating hypothetical treatments and creating ideas for experimentation and refinement. However, the internal reality of an AI model rarely corresponds exactly or even at all to real world situations. It's only offering a theory based on a simulative model that can range from somewhat accurate to totally wrong.
For example, I used Gemini 2.0 Flash to try to come up with solutions to blend a patch of new paint into the existing paint of a damaged wall. A lot of its solutions were not effective because its internal model of how paint functions was not always complete or accurate.
It suggested paint matching apps that failed to identify blue as blue. It suggested a paint strategy that didn't account for the speed of paint drying. It suggested thinner coats of paint at the edges of the damage to blend into the older paint; this instead created a border of darker paint at the perimeter. It suggested paint matching techniques that didn't account for sheen. It suggested blending mismatched paint by diluting the paint with acrylic gloss to have it gradually lighten at the edges and this last one actually worked.
Where AI can be helpful: I struggle at coming up with ideas to test. AI can generate ideas faster so that I can focus on testing them and refining the ones that don't immediately fail. It speeds up the process of trial and error by offering ideas for trial. However, its hypotheticals are not solutions; they're only ever starting points for tests.
]]>Running a vending machine requires analyzing purchasing patterns and making decisions on inventory costs, setting sale prices and markup, engaging with customer needs, and responding to sales patterns. AI failed to do this with anything even vaguely resembling coherence.
Generating a pair of two-second special effects shots showing a collapsing building? That's simply a matter of creating 24 frames per second, and in this case as straightforward as creating 96 images and as a variation on one core template image. It is a much simpler task. To compare the two is like comparing brain surgery to drawing a picture of a band-aid.
Certainly wasn't comparing the two, just showing how AI continues to integrate into society, like Netflix adopting it. Although in time, I do believe it will eventually be able to do those tasks for vending machines. Elon and Sam Altman keep doubling down on AI, but they're gonna need a lot more electricity. On the other hand I wonder if they will even need it if, and that's an if, it can figure out equations we haven't solved or even discovered. AI should be able to help me in the long run. I can't drive due to my illness, and perhaps it can find new medications and solutions that could help me. I do find ChatGPT and Grok helpful with info of my health and giving me some ideas. Nevertheless, I'm looking forward to Optimus murdering me in my sleep.
]]>Generating a pair of two-second special effects shots showing a collapsing building? That's simply a matter of creating 24 frames per second, and in this case as straightforward as creating 96 images and as a variation on one core template image. It is a much simpler task. To compare the two is like comparing brain surgery to drawing a picture of a band-aid.
]]>AI is given a single vending machine to run, quickly runs its own business into the ground.
https://www.inc.com/ben-sherry/an-ai-ra … e/91207636
Meanwhile,
'We remain convinced that AI represents an incredible opportunity': Netflix used AI special effects in original series instead of paying VFX artists
"Using AI powered tools, they were able to achieve an amazing result with remarkable speed," Netflix co-CEO Ted Sarandos said in a July 17 earnings call.
]]>AI is given a single vending machine to run, quickly runs its own business into the ground.
https://www.inc.com/ben-sherry/an-ai-ra … e/91207636
wow
]]>