What most companies are selling as AI is just a lie.
Here’s a great vid talking about what can be useful about AI vs how most people try to use it.
1 Like
https://twitter.com/samuelwoods_/status/1642889718336479233
I really want to use this one:
https://twitter.com/varunshenoy_/status/1640764666946424832
Having an editor for any English text that I write would be terrific. I could run just about anything through there, even just comments on web sites.
1 Like
ChatGPT can make piracy easier.
EDIT: Similarly, this too. lol
Someone won the 2023 SONY World Photography award with an AI generated image. They weren’t trying to cheat, they were trying to prove that the photo contests were woefully unprepared. They declined to accept the award or the prize.
2 Likes
Yes, AI WILL destroy us.
Trash tabloid magazines taking advantage.
At least for now, filtering out garbage tier content created by ML still isn’t any harder than filtering out garbage tier content created by human hands.
And if the ML does create some content that isn’t garbage, does it matter how it was made if it has such value?
Huh. That’s actually extremely valuable data.
2 Likes
AI surveillance is about to get worse.
The copy paste algorithm knows how to copy paste. While is not a strictly verbatim example, the same quality of results are just a google search away. What’s the difference?
1 Like
Worse than googling in every respect!
IBM is now replacing jorbs with compooters.
1 Like
YouTube suggested this video to me for obvious reasons, and it is really something else.
Someone took the recorded singing voices of the original 9 members of Girls’ Generation and trained a generative AI model for each one. Then they took the very popular song “Ditto” by New Jeans, split up the parts, and had the artificial GG perform a cover.
As someone who has listened to a lot of Girls’ Generation music, only Yoona’s voice is off. Not a surprise since she doesn’t sing a lot of lines, so the model had less data to work with.
First of all, I think this might be the end, or really a new beginning, for Hatsune Miku’s Vocaloid technology. The way the Vocaloid software works now is that it’s very similar to a MIDI instrument. You put in notes and syllables, and it sings. It’s clear that a generative AI model can just do the same thing only better. Train it in the voice you want it to sing with, hand it the song, and it’s all good.
Also, if this technology is now so easy to use that some 1.3K subscriber YouTuber can throw them together, oh boy. I think it’s going to blow up.
Lastly, this one was pretty easy to get a good result with because it’s a KPop girl group covering another KPop girl group. What would happen if I asked it to do something that was really dissimilar. Like can I get AI Lil’ Jon do a cover of Nat King Cole or vice versa?
1 Like
We’re getting real close to being able to train the bots to make anyone cover whatever to an acceptable level.
There’s a lot of examples of things like this:
which are a lot closer than the similar examples from a few months back.