The point is this theory starts from the premise that LLMs have knowledge and the capacity to reason to prove that they have knowledge and the capacity to reason. This is not how you prove anything. Their experiment literally has the LLM evaluating its own output. This is just a slightly less cringe version of that cajun discordian google employee.
on one hand, lmao my day job is supporting developers so Copilot falls squarely in our wheelhouse. my boss has been actively looking for people interested in trialling it to see if itâs worth subscribing to. I think itâs a load of crap
on the other hand, I canât share this article else look like a sneering stick-in-the-mud, back in my day we wrote bash scripts by hand both ways
i have cemented my place as a sneering stick in the mud, personally, and people have responded surprisingly well. turns out a lot of people feel similarly: skeptical but scared of being perceived as against change.
Was there ever a resolution to the whole âthis code you trained your ai with is under a copyright licenseâ problem? I thought that would nip this whole fad in the bud
I have A Reputation but also get a lot of traction discussing this and Ludditism with one of the data scientists who is more than happy to argue against AI
Shocking! I wonder what the next big grift will be. Maybe quantum computing?
not obviously socially corrosive enough. iâm thinking autonomous home defense systems
or fully automated electronic managers making inroads into more industries, maybe like a turnkey version of what amazon workers have to put up with?
Maybe they can go full Oryx and Crake and get really into monetizing genotypes and doing weird patented gene splicing. You know, try to subsume as much of the medical industry into the tech industry as possible.
âThe Internet of Beingsâ
Start with pets. Designer pets are a lot more palatable to the public than designer humans, which you can always introduce later on down the line after people get used to designing their next dog or cat.
we already have literal âdesigner breedsâ
and yet no one is charging licensing fees for their genes. money on the table!
The event I had to work today was a local film maker showing off the first chunk of the documentary hes trying to crowd fund. He had some good interviews but on top of one of the intervieweeâs audio being totally out of sync like 80% of what was on screen were photograph style AI art of historical scenes, primarily of indigenous people. I know first hand finding available and good visuals that themselves arent iffy sketches and paintings that were made by Europeans is pretty damn hard but phoneying up images a population that has a long history of being misrepresented aint the way to do it. Plus Iâm sure whatever they were using was trained on like bootleg Harley Davidson shirts off of eBay and Daniel Day Lewis.
This shit drives me nuts, a bunch of YT video essayists have started being like âoh, I can just generate relevant looking AI art to throw up on the screen while I talk about stuffâ and it always looks like complete garbage, especially if theyâre talking about historical events because theyâll just generate art based on paintings/drawings of that time period only because itâs AI, itâll inevitably throw in a really obvious anachronism that nobody catches because nobody cares. Just the absolute laziest bone-stupid approach, âooooh itâs free contentâ, youâd be better off with just a solid color background youâre talking over, you dunces.
yeah i hate this, i unsubscribed from a channel that posts dwarf fortress gameplay footage because of the gross ai art thumbnails