4 ms·

> The minimal much easier to reach requirement it needs to do that is make it so people can't find jobs How tho? Will LLMs replace grocery store workers? Farme

> The minimal much easier to reach requirement it needs to do that is make it so people can't find jobs How tho? Will LLMs replace grocery store workers? Farmers? Carpenters? Plumbers? Electricians? Truck drivers? Train drivers? Chefs? Cleaners? They won't even "replace" middle management. Yes, they can help them write emails, plan stuff, research it and do all the fluff boilerplate parts of the work. Can it communicate with people properly, understand when somebody is overestimating, underestimating, lying, not lying, when the worker is actually sick or just pretending? In the twitter-hn echo chamber, it seems like an LLM can do it all since it can "write text and code", but literally nobody doing any work outside of "we write text" sphere isn't really worried. Hell, as a developer I ain't even blinking - I've worked with GPT for more than a year and am well aware of it's limitations - and yes, some can be solved by chaining another instance to check the instance and then another instance to whip the instance into submission, but that is just unproductive and way more costly to execute than just to have a human expert do it. Also, innovations that increase productivity don't lead to losing jobs, they lead to increased productivity and economic value. Otherwise, our civilisation would be starving since the wheel. It's more likely we will slowly progress towards the "universal income" utopia by using NGAI (near-general AI) than we will devolve into a bloody chaotic hungry mess. > because companies don't buy the labor But who will buy those companies products? What do you define as labor? As far as I know, OpenAI isn't building highly advanced robots with multi-sensory functionality that are as robust and as adaptable as humans. If they were, I'd be a bit more afraid. Just a bit.

_siis 4y ago
The how is unimportant, focusing on it is a forest for the trees blunder. It can happen any number of different ways, and you won't know which way until its too late to do anything about it. The danger of cascade failures is you don't perceive an issue until it is too late to do anything about it, most times. The rate of change is very important. ~10 years ago, object tracking and detection couldn't detect objects very accurately. This was before resnet. Then resnet came out and it changed everything. Then NLPs, Then Visual Transformers. 3dimensional depth mapping, GANs and now LLMs. Its progressive and exponential. No one could predict how soon the next breakthrough would be but they fell like dominoes so far at an increasing rate. You don't need middle management if you have no low level employees. Self-service kiosks are already being used in a lot of countries to replace workers. 3D printing is replacing other areas. They already have cleaning robots. Robots are fixed cost assets. People's expertise are recurring cost assets. Its clear which is the better investment from a business perspective short-term. Anyone that isn't really worried about these things is going to be blindsided because they fundamentally don't understand the nature of the problem. Imagine a Code-GPT comes out in a few months that reads a requirements specification written in regular language and outputs code you'd normally have spent months writing in a few seconds. No expertise needed. You won't know ahead of time. Someone does it, publishes it, your entire profession is mostly out of a job. If multiple chains cost cents, how much chaining needs to happen for it to equal minimum wage, let alone a mid-level salary? You are wrong about your statement that innovations that increase productivity don't lead to losing jobs. They do when those productivity gains exceeds demand. Demand is constrained by the population/market size. While I'm not too worried about losing my job since I tend to do high IQ design work, those low and mid-IQ jobs normally filled by average or above average intelligence people will be hit the hardest. You should be afraid, simply because the rate of change exceeds our ability to react in any meaningful way. If there is a problem, we'll never be able to see it coming until it is too late. Also, when has there never been some problem. If you've done any significant study on socialism, you'd know both market and non-market socialism suffer from the economic calculation problem. Its not solved, people have been working on it for over a hundred years. Shortages cause failures which then cause death on a large scale. Universal Income is equivalent to non-market socialism where resources are simply moved around internally. There is no actual price discovery, prices deviate so much from the rational cost to produce those items that you run into characteristic issues of this problem. Economic value by itself offers little. It can be solely concentrated in one person, or diffused throughout an entire country. One has benefit the other doesn't, but in terms of economic value they are the same. Its a made up number that doesn't measure what it claims to.

_siis 4y ago

The how is unimportant, focusing on it is a forest for the trees blunder. It can happen any number of different ways, and you won't know which way until its too late to do anything about it. The danger of cascade failures is you don't perceive an issue until it is too late to do anything about it, most times. The rate of change is very important. ~10 years ago, object tracking and detection couldn't detect objects very accurately. This was before resnet. Then resnet came out and it changed everything. Then NLPs, Then Visual Transformers. 3dimensional depth mapping, GANs and now LLMs. Its progressive and exponential. No one could predict how soon the next breakthrough would be but they fell like dominoes so far at an increasing rate. You don't need middle management if you have no low level employees. Self-service kiosks are already being used in a lot of countries to replace workers. 3D printing is replacing other areas. They already have cleaning robots. Robots are fixed cost assets. People's expertise are recurring cost assets. Its clear which is the better investment from a business perspective short-term. Anyone that isn't really worried about these things is going to be blindsided because they fundamentally don't understand the nature of the problem. Imagine a Code-GPT comes out in a few months that reads a requirements specification written in regular language and outputs code you'd normally have spent months writing in a few seconds. No expertise needed. You won't know ahead of time. Someone does it, publishes it, your entire profession is mostly out of a job. If multiple chains cost cents, how much chaining needs to happen for it to equal minimum wage, let alone a mid-level salary? You are wrong about your statement that innovations that increase productivity don't lead to losing jobs. They do when those productivity gains exceeds demand. Demand is constrained by the population/market size. While I'm not too worried about losing my job since I tend to do high IQ design work, those low and mid-IQ jobs normally filled by average or above average intelligence people will be hit the hardest. You should be afraid, simply because the rate of change exceeds our ability to react in any meaningful way. If there is a problem, we'll never be able to see it coming until it is too late. Also, when has there never been some problem. If you've done any significant study on socialism, you'd know both market and non-market socialism suffer from the economic calculation problem. Its not solved, people have been working on it for over a hundred years. Shortages cause failures which then cause death on a large scale. Universal Income is equivalent to non-market socialism where resources are simply moved around internally. There is no actual price discovery, prices deviate so much from the rational cost to produce those items that you run into characteristic issues of this problem. Economic value by itself offers little. It can be solely concentrated in one person, or diffused throughout an entire country. One has benefit the other doesn't, but in terms of economic value they are the same. Its a made up number that doesn't measure what it claims to.