This article highlights a number of the LLM failures in the real world. Hiltzik: The Disney/Sora fiasco shows the limits of the AI craze
8 Likes
Looking at the statistics we have available I think we are getting close to the AI bubble popping. But unlike what many think AI will stay a thing. Like think of the .com bubble. It just gets put into its natural place. I think that’s a good thing all in all. Ai has good uses for coding. I’ve used AI to find bugs I was stuck on for months. But all the issues with AI right now are waaaaaay too big.
2 Likes
Interesting read .
Article shown how companies reacted even when “AI” caused issue .
For movie industry this is determined how Netflix will continue use it or will return to more traditional scripts , voiceovers .
Disney betrayal is nicest example there , but what is it means if they still use it ? 
For Amazon ? Microslop ? - does consumers discourage them from keep implementing it 
The biggest thing might be if we experience another stock crash and Nvidia would go with it , this could cause chain reaction..
I think hating on LLMs is a mistake.
You can hate what companies did to train their LLMs.
You can hate what corporations will do with LLMs to replace their staffing costs.
You can hate what people do with LLMs.
You can hate the environmental impact of LLMs.
But to hate LLMs feels a bit like hating calculators.
They are amazing. They will continue to evolve the way we interact with machines. Yes governments will use them for spying, yes greedy rich will use them to cut costs to increase profits. Yes governments should not be scared to say no to resource draining massive data centres. But this is humans being human. Yes we are short sighted, selfish and greedy. Our culture seems to be heading in directions we don’t want it to go, like monitoring, observation, and control. We seem to be entering a time of the powerful super rich, some of whom have more money than entire countries. The system that controls our culture seems broken.
But even given all that, LLMs are not evil nor good. They are a tool. You may as well feel hatred towards a hammer.
The only thing we can do is to refuse to pay for it directly, and where possible indirectly too with our votes and our attention. Personally I had always hoped that the AI slop generated on social media sites would be their ultimate downfall. When finally no humans are using it, who is left to pay for anything?
Anyway, the idea of Disney being a force for wonder and good, was long ago dispelled as they are revealed as just another grubby profit-greedy machine spreading propaganda. I hope the disney parks go bankrupt too. Feel free to hate on Disney, and OpenAI, but not LLMs themselves.
6 Likes
Judging by how many people prefer actual junk food slop to real food, I think it’s safe to assume that there will always be people who prefer convenience to quality. Hell, I’ll probably be sprinkling in some of that refined AI sugar myself, when I need to remove the backgrounds from 100 PNG’s or whatever, just like I put sugar on my corn flakes 
Right now I think people are starting to see that there are very few fields where AI can actually compete in terms of quality - it’s just able to do better than a layperson, so 99% of us are impressed when an AI can spit out a horribly derivative song that would make a professional musician want to vomit.
From a business perspective, it’s just about walking that edge and figuring out where quality is important (ie. utilities customer service) and where it isn’t (ie. auto-generating thumbnails). Like the oil industry, that’s not going to end before it’s impossible to squeeze out any profit at all.
2 Likes
I admit that was a clickbait topic.
You can’t separate out how the tool is made and used, the people who use it and the way they use it, and then put the tool on its own blameless pedestal. Sure, LLMs have no will of their own, but how they affect the world matters.
Comparing LLMs to calculators or hammers is what’s know as a false equivalence logical fallacy.
Hammers are not made with a disproportionate number of resources to all comparable tools. They are not made by stealing the work of carpenters and not compensating them for it. They have not caused layoffs because a hammer can do the job of 10 carpenters. They are not used to replace screwdrivers.
I think this is a simplification of the fact that a lot of people can only afford junk food. If you look at the hoops fast food companies are jumping through to keep poor people buying food, you can see how they are not because it’s too expensive now. So I think it’s least common denominator.
2 Likes
Perhaps! I don’t live in the United States, so that scenario is unfamiliar to me In Europe, raw ingredients are generally cheaper than processed ones, and yet there’s still a huge market for junk food.
My point isn’t about food though - what I’m saying is that convenience and quality are often at odds, and people the world over tend toward convenience to some degree. Of course some people will prefer the highest quality, regardless of convenience - say, people who import coffee beans, toast them themselves, grind them themselves and use an expensive machine to make 3 cl of espresso. However the majority are satisfied with an easier, cheaper, faster solution, and AI is the freeze-dried, supermarket-bought coffee product of data processing.
3 Likes
Yeah, personally i think AI would be better if it was only trained on each person’s own stuff, but then that would create more ethical risks. . . heck, I cant find a good excuse for AI.
AI is a lot more complex than a calculator, and is also a lot more dangerous.
i would have said “Hating on AI is like hating on factories”, this is basically a Industrial Revolution repeat minus the fact that the AI bubble is gonna burst soon.
(As for Disney and OpenAI, i cant stand either, Microsoft is just as bad if not worse. )
3 Likes
There be always good and bad side .
Ethically if you use Godot docs , engine code and demo projects to make GodotAI LLM would this be ok ?
It would be legal, but not necessarily ethical. Because you didn’t ask the maintainers of the Godot project for permission.
6 Likes
This, but I’d much rather have this happen than what already happened with Meta pirating terabytes of books and getting away with it because it’s Meta.
If I tried that I’d be in jail for the rest of my life. Yes very fair. Thank you big corporations.
3 Likes
LLMs suck because of their nature. They steal and blur together people’s art and writing without consent and aren’t even perfectly accurate
2 Likes
except the factory workers who get replaced love their jobs and do it better
Technically it’s more like the Second Industrial Revolution which started in the 1870s and went until 1914 when the first World War started. Between the World War and industrialization which took about 50 years, it led to the Great Depression. And in that way, the AI Bubble burst is looking a lot more like a combination of the DotCom Bubble and the Great Depression rolled together.
And people hated the factories because they lowered wages on skilled labor because they were replaced by a machine that could do that labor with an unskilled worker cheaper and faster. Sounds a lot like LLMs, doesn’t it? So I’d agree that’s a good comparison.
Have you ever lived in a factory town? They do not love their jobs. They love that it means they can support their families and do what they want with their free time. Sadly, advancements mean that they are being retrained all the time because frankly, robots can do parts of their job better. They get upset because losing their job means taking a job that pays less somewhere else and can cause a loss of quality of life.
3 Likes
(post deleted by author)
1 Like
thats why I said “except”
last time I checked factory bots dont steal gallons of water and millions of pieces of artwork without consent to do their work
There is an additional problem showing up. The data centers will heat up the surface around them. Not just a bit.
https://archive.fo/Riztz
Then I am misunderstanding your usage of the word in that context, because I am used to it being used to accentuate a point, not counter one with that usage.