Sign In
The great A.I lie | Forum
Support Dantalion
Sep 13 '25
So I’ve read tons of articles and posts from Silicon Valley A.I entrepreneurs who say that A.I is going to do all of our jobs, and people won’t have to work. I even read one where a guy said that we will get paid 10,000 dollars a month to stay home and do nothing. You’ve got to be fucking stupid to believe that. Where is this money going to come from, the very companies that are going to fire us and replace us with robots? It sure as fuck isn’t going to come from the government. The countries that will be mass adopting A.I are deep into debt. If anything, governments will be using increased revenue from A.I to pay down debt, not give to the people. And if you think that there is going to be handouts from companies when you don’t work for them, you’re dead wrong. Why would a company replace all their workers with A.I if they aren’t going to save on labor costs and be more efficient. Here’s what’s going to really happen. At least in the U.S, we will have a 20,30, maybe even 40% unemployment rate, and those people will be forced to do any side hustle they can to survive that doesn’t involve getting a paycheck from a company. You’ll have lots of backyard mechanics, junk removers, babysitters, domestic caregivers….all low wage with no benefits or stability. Social Security will collapse. Along with other safety net programs. Yet you’ll have about 7,000 billionaires and a few trillionaires. That’s what’s going to happen. You’ll still have to hustle (probably harder) just to survive, while the mega rich buy 50 mansions and 10 yachts. Nobody is going to give you shit.
Brother_Telluric
Sep 16 '25
Brother Dantalion, it can make room for new progress, exactly that relese of workforse, can enable it. Similar to industrial revolution maybe. I agree there will be job losses but in lower percentage, i do not see why however would then service jobs , utility, care givers, servicemen/repairman, police , military, religion?, turism, ect would be low payed . Completley automated factories where reality in Italy for example in early 90s Videocolor carhode ray tube production facility had workers who just packed finished product ,and dudes feeding raw materials.. So concept is tested decades ago in a sense that you can have enterprise that is viable. It could be many such enterprises creating stuff and then you have big consumer base for growth. It can even lead to deflation and strengthen the currency. Japan is experimenting with artificially downgrading automation potential in society it already possess, and it works(somewhat). We will see, but to me it looks like a semi inflated baloon(current state of AI affairs )that will pop but not completely , simmilar, yet different than dot com bubble.
The Forum post is edited by Brother_Telluric Sep 16 '25
Brother_Telluric
Sep 16 '25
And i got it more jobles people, lower the pay, BUT assuming there will be some rock solid agreement of guarantees, that will strengthen the currency redistribute /enable cash flow, real value redistribution on structural level, it could be that strengthening the currency IS possible
Jupiter Wyrd
Nov 19 '25
I don't believe the hype about AI replacing our jobs, much less humanity, in the next few years, what I do believe is that this type of narrative originates from tech CEOs in order to get more and more investments.
There are a few major issues with the current AI models in use, the main models being language-based models and image-based ones. These run into issues such as prompt injection (getting an AI to forget previous instructions in favor of a new one, inputted by the user (note that I said the user and not the developper - LLMs cannot distinguish between what devs teach it and what users teach it, since both are based in language)). LLMs are very vulnerable to this. Another major issue is that our current AI models can't extrapolate, only interpolate; they cannot come up with anything that doesn't already exist in the data sets they look at. And since these data sets are either language or image-based, it is easy for them to 'hallucinate' an answer to a question it does not have clear data on, because it cannot correlate it with any other form of data.
Current models have security issues (prompt injection) which cannot be solved fundamentally because they are rooted in the nature of the model - language being easy to manipulate. They cannot become smarter than the existing material (data) they look at. To create real AGI, you would need models that are very different, possibly crossing different existing models and some non-yet-existant ones. Currently, the hype is creating a financial bubble, I think.
An anecdote is that I tried using AI (GPT and CoPilot) to help me design a small video game project. There was an issue with my GUI that the AI just could not figure out (neither could I, as I was new to the engine I was using and to game dev in general), and no matter how much info I sent it, it could not solve the issue. If I had gotten a human (who knew the game engine well) to look at it, I'm sure they could have pinpointed where the problem was in two seconds. Granted, I was using a free version of these tools. Nonetheless, I was not impressed at all. To get an AI to solve a specific type of problem, it would have to be specialized in solving that kind of problem. But then that is no longer an AGI, is it?
I think someday, there may be AI powerful enough to replace all jobs, and even humanity. I don't know. It's hard to say what the future will look like. I do not think current AI models are anywhere close to doing these things, though.
There are a few major issues with the current AI models in use, the main models being language-based models and image-based ones. These run into issues such as prompt injection (getting an AI to forget previous instructions in favor of a new one, inputted by the user (note that I said the user and not the developper - LLMs cannot distinguish between what devs teach it and what users teach it, since both are based in language)). LLMs are very vulnerable to this. Another major issue is that our current AI models can't extrapolate, only interpolate; they cannot come up with anything that doesn't already exist in the data sets they look at. And since these data sets are either language or image-based, it is easy for them to 'hallucinate' an answer to a question it does not have clear data on, because it cannot correlate it with any other form of data.
Current models have security issues (prompt injection) which cannot be solved fundamentally because they are rooted in the nature of the model - language being easy to manipulate. They cannot become smarter than the existing material (data) they look at. To create real AGI, you would need models that are very different, possibly crossing different existing models and some non-yet-existant ones. Currently, the hype is creating a financial bubble, I think.
An anecdote is that I tried using AI (GPT and CoPilot) to help me design a small video game project. There was an issue with my GUI that the AI just could not figure out (neither could I, as I was new to the engine I was using and to game dev in general), and no matter how much info I sent it, it could not solve the issue. If I had gotten a human (who knew the game engine well) to look at it, I'm sure they could have pinpointed where the problem was in two seconds. Granted, I was using a free version of these tools. Nonetheless, I was not impressed at all. To get an AI to solve a specific type of problem, it would have to be specialized in solving that kind of problem. But then that is no longer an AGI, is it?
I think someday, there may be AI powerful enough to replace all jobs, and even humanity. I don't know. It's hard to say what the future will look like. I do not think current AI models are anywhere close to doing these things, though.
The Forum post is edited by Jupiter Wyrd Nov 19 '25
JosephRose
Feb 4
AI is a tool, and a useful one that will become more useful in time. There will be job displacement, as with any technological changes that happen.
