Short term yes; long term probably not. All the dipshit c-suites pushing the “AI” worker replacement initiatives are going to destroy their workforces and then realize that LLMs can’t actually reliably replace any of the workers they fired. And I love that for management.
Lol AI cannot replace either of those jobs. “I’m sorry I can’t help with your time off request but here is a gluten free recipe for a pie that feeds 30 people.”
You’re referring to something that is changing and getting better constantly. In the long term LLMs are going to be even better than they are now. It’s ridiculous to think that it won’t be able to replace any of the workers that were fired. LLMs are going to allow 1 person to do the job of multiple people. Will it replace all people? No. But even if it allows 1 person to do the job of 2 people, that’s 50% of the workforce unemployed. This isn’t even mentioning how good robotics have gotten over the past 10 years.
Sure you sort of need that at the moment (not actually everything, but I get your hyperbole), but you seem to be working under the assumption that LLMs are not going to improve beyond what they are now. It is still very much in its infancy, and as the tech matures this will be less and less until it only requires few people to manage LLMs that solve the tasks of a much larger work force.
It’s hard to improve when the data in is human and the data out cannot be error checked back against its own data in. It’s like trying to solve a math problem with two calculators that both think 2+2 = 6 because the data they were given said that it’s true.
LLMs now are trained on data generated by other LLMs. If you look at the “writing prompt” stuff 90% is machine generated (or so bad that I assume it’s machine generated) and that’s the data that is being bought right now.
There is a plateau to be hit at some point. How close it is, depends who you ask. Some say we are close, others say we are not but it definitely exists. LLMs suffer, just like other forms of machine learning, from data overload. You simply can’t be infinitely feeding it data and keep getting better and better results. ChatGPT’s models got famous because value function for learning had humans involved who helped curate quality of responses.
It can potentially allow 1 worker to do the job of 10. For 9 of those workers, they have been replaced. I don’t think they will care that much for the nuance that they technically weren’t replaced by AI, but by 1 co-worker who is using AI to be more efficient.
That doesn’t necessarily mean that we won’t have enough jobs any more, because when in human history have we ever become more efficient and said “ok, good enough, let’s just coast now”? We will just increase the ambition and scope of what we will build, which will require more workers working more efficiently.
But that still really sucks because it’s not going to be the same exact jobs and it will require re-training. These disruptions are becoming more frequent in human history and it is exhausting.
We still need to spread these gains so we can all do less and also help those whose lives have been disrupted. Unfortunately that doesn’t come for free. When workers got the 40 hour work week it was taken by force.
My colleagues are starting to use AI, it just makes their code worse and harder to review. I honestly can’t imagine that changing, AI doesn’t actually understand anything.
This comment has similar vibes to a boomer in the 80s saying that the Internet is useless and full of nothing but nerds arguing on forums, and he doesn’t see that changing.
Probably. I’m just not seeing it actually doing any logic or problem solving. It’s a pattern matching machine today. A new technology could certainly happen.
Do you know what pattern matching is great for? Finding commonly cited patterns in long debug log messages. LLMs are great for brainstorming problem solving. They’re basically word granularity search engines, so they’re great for looking things up are more niche knowledge that document search engines fail on. If the thing you’re trying to look up doesn’t exist, it will make shit up so you need to cross reference everything, but it’s still incredibly helpful. Pattern matching is also great for boilerplate. I use the codium extension and it comes up with auto complete suggestions that don’t have much logic, but save a good amount of key strokes.
I didn’t think the foundational tech of LLMs are going to get substantially better, but we will develop programming patterns that make them more robust and reliable.
That’s not what is going to happen. Copilot will simply increase productivity over, and where before they needed 10 people, gradually, through attrition they will need only 9, then 8, and so on. That does not mean higher unemployment though, it means more product.
Businesses want to grow, not keep stable. They might fire a few ppl in the short term, but in the long term it’s more likely the group of 10 would just do now the work of a 12-13 group with AI, producing hugher outputs for the same money they were getting before, meaning extra profit for the shareholders.
Short term yes; long term probably not. All the dipshit c-suites pushing the “AI” worker replacement initiatives are going to destroy their workforces and then realize that LLMs can’t actually reliably replace any of the workers they fired. And I love that for management.
They’re gonna realize the two jobs it can actually replace is HR and the C suite.
And neither of those two groups will allow themselves to be replaced.
Yeah, HR gets by because of legal compliance, and execs get by through convincing the board to give them X years, and then jump to the next one.
Lol AI cannot replace either of those jobs. “I’m sorry I can’t help with your time off request but here is a gluten free recipe for a pie that feeds 30 people.”
You’re right, that sounds better than the average HR rep.
And better than the average pie 😋
“Help” with a time-off request?
Here’s the help:
It won’t replace any jobs entirely, it will just reduce the number of people needed for each job.
Not that there’s much difference if you’re the one being made redundant.
Well at least you’d get a recipe
I’d unironically like that recipe, please
Ingredients: 1 potato
Steps:
Cut into 30 prices and serve
I bet project managers could be replaced with AI super easily, I mean all they have to do is respond to all messages with 👍
Then you don’t have good project managers.
At least then the project plan would get updated and tasks opened on time…
You’re referring to something that is changing and getting better constantly. In the long term LLMs are going to be even better than they are now. It’s ridiculous to think that it won’t be able to replace any of the workers that were fired. LLMs are going to allow 1 person to do the job of multiple people. Will it replace all people? No. But even if it allows 1 person to do the job of 2 people, that’s 50% of the workforce unemployed. This isn’t even mentioning how good robotics have gotten over the past 10 years.
You must have one person constantly checking for hallucinations in everything that is generated: how is that going to be faster?
Sure you sort of need that at the moment (not actually everything, but I get your hyperbole), but you seem to be working under the assumption that LLMs are not going to improve beyond what they are now. It is still very much in its infancy, and as the tech matures this will be less and less until it only requires few people to manage LLMs that solve the tasks of a much larger work force.
It’s hard to improve when the data in is human and the data out cannot be error checked back against its own data in. It’s like trying to solve a math problem with two calculators that both think 2+2 = 6 because the data they were given said that it’s true.
How is it hyperbole? All artificial neural networks have “hallucinations”, no matter their size. What’s your magic way of knowing when that happens?
LLMs now are trained on data generated by other LLMs. If you look at the “writing prompt” stuff 90% is machine generated (or so bad that I assume it’s machine generated) and that’s the data that is being bought right now.
There is a plateau to be hit at some point. How close it is, depends who you ask. Some say we are close, others say we are not but it definitely exists. LLMs suffer, just like other forms of machine learning, from data overload. You simply can’t be infinitely feeding it data and keep getting better and better results. ChatGPT’s models got famous because value function for learning had humans involved who helped curate quality of responses.
It can potentially allow 1 worker to do the job of 10. For 9 of those workers, they have been replaced. I don’t think they will care that much for the nuance that they technically weren’t replaced by AI, but by 1 co-worker who is using AI to be more efficient.
That doesn’t necessarily mean that we won’t have enough jobs any more, because when in human history have we ever become more efficient and said “ok, good enough, let’s just coast now”? We will just increase the ambition and scope of what we will build, which will require more workers working more efficiently.
But that still really sucks because it’s not going to be the same exact jobs and it will require re-training. These disruptions are becoming more frequent in human history and it is exhausting.
We still need to spread these gains so we can all do less and also help those whose lives have been disrupted. Unfortunately that doesn’t come for free. When workers got the 40 hour work week it was taken by force.
My colleagues are starting to use AI, it just makes their code worse and harder to review. I honestly can’t imagine that changing, AI doesn’t actually understand anything.
This comment has similar vibes to a boomer in the 80s saying that the Internet is useless and full of nothing but nerds arguing on forums, and he doesn’t see that changing.
Probably. I’m just not seeing it actually doing any logic or problem solving. It’s a pattern matching machine today. A new technology could certainly happen.
Do you know what pattern matching is great for? Finding commonly cited patterns in long debug log messages. LLMs are great for brainstorming problem solving. They’re basically word granularity search engines, so they’re great for looking things up are more niche knowledge that document search engines fail on. If the thing you’re trying to look up doesn’t exist, it will make shit up so you need to cross reference everything, but it’s still incredibly helpful. Pattern matching is also great for boilerplate. I use the codium extension and it comes up with auto complete suggestions that don’t have much logic, but save a good amount of key strokes.
I didn’t think the foundational tech of LLMs are going to get substantially better, but we will develop programming patterns that make them more robust and reliable.
deleted by creator
Short term? Sure.
Long term? Not a chance that equation works out favorably.
But then again, c-suites these days only seem to give a shit about short-term implications.
That’s not what is going to happen. Copilot will simply increase productivity over, and where before they needed 10 people, gradually, through attrition they will need only 9, then 8, and so on. That does not mean higher unemployment though, it means more product.
“AI means there will be fewer people required to do the same amount of work”
“this does not mean higher unemployment”
I think you left out a steep off reasoning there. At least, I don’t follow.
When productivity increases (as it has been doing for ages) the manufacturing output increases. That’s what normally happens.
But the amount of workers will only stay the same if demand grows at the same rate as the production output.
Well, the price goes down, or/end the salaries go up, or resources are freed for new investments…
Only in the last case there is a chance that the amount of jobs will remain the same, the other cases will lead to lost jobs.
Prices going down leads to increased demand and expansion. Salaries (everywhere) going up lead to increased demand and expansion.
Businesses want to grow, not keep stable. They might fire a few ppl in the short term, but in the long term it’s more likely the group of 10 would just do now the work of a 12-13 group with AI, producing hugher outputs for the same money they were getting before, meaning extra profit for the shareholders.
That’s exactly what I meant by