All of which comes from an era without the public internet, the smartphone, or AI.
A famous catchphrase, this time it will be different. Hope is not a strategy.
If we're trading catchphrases, here's one I prefer:
"In times of change learners inherit the earth while the learned find themselves beautifully equipped to deal with a world that no longer exists." - Eric Hoffer.
That said, I'd prefer it if someone addressed the point, which is that the analysis of what's possible dates from an earlier technological era.
If the contention is that technology doesn't change what's possible in society and that society will always be the way it is was for most of the 20th century, I think we can safely ignore that.
The contention is if history has gone through multiple and innumerable cycles and has ended up at the same outcome multiple times before, given an infinite number of potential outcomes, there must be common principles that led to the clustering at that outcome.
If you choose to ignore what those conveniently documented outcomes were and by extension what led to them, can you exercise any reasonable control over something you have no knowledge of?
Would the statement, "This time will be different, have any credibility if you have no control or knowledge of the issues?
Everyone at some point will say anyone, anything, and nothing, and what is said will likely be less true than not true when it is not backed with support.
Rational and rigorous approaches from first principles lend credibility to something being potentially more true than not, and by the law of approximation we can improve these approches over time until we eventually get to the point of it being true.
Wouldn't you say its better to be true, than not true if you want to succeed and live?
The first problem is that this is simply a false premise. Very few people would make such a claim, and none can back it up with evidence.
The second is that you still seem to think that technological change has no impact on history.
To claim otherwise is simply discounting a large body of scientific work which is highly credible and supported, how can you claim doing so is credible in any way?
The second is just putting words in my mouth, I have no idea how you could possibly come to that conclusion about what I think based off our short conversation here.
I never said it, I never inferred it, it looks to me simply like a psychological projection you've made of something you think, Largely because I don't believe that in the slightest. Not in the slightest, and no rational person would.
So can you clarify 'specifically' what I said that made you think that?
Did these past catastrophes involve AI and computers?
It reminds me of Y2K hysteria.
When old jobs die, new jobs arise. That has always been the case, and I haven't seen any evidence of that ending.
The problem is that humans aren't getting specially smarter as generations go, but technology continue to grow
It also ignores a growing body of evidence.
If new jobs arise as you say in the same amounts, why is it that young people are having such a hard time finding gainful employment these days?
Take a look at the statistics, you'll find as some already have, that the job pool has been shrinking, and older workers who have more experience out-compete younger workers for entry level jobs because the cost to train (is high), and production is lower in workers with no experience than those with experience.
The only indicators we would have that this is happening, just like inflation, are lagging indicators.
By the time you know it is happening, its then too late to do anything. That's why this problem will probably kill us all. We as a species historically do not handle cascade failures well because we don't perceive the danger until its too late.
Its a problem where our natural biases work against us. Its normal to just discount problems we have no context, which is why the out-of-context problem is so dangerous.
Do you think the native people's of the Carribean thought there was danger from white sails of Columbus ships?
By what measure? Youth unemployment is right now around its lowest level in…quite a long time. [0]
[0] https://tradingeconomics.com/united-states/youth-unemploymen...
BLS data has been manipulated and methodology changed to always paint the rosiest picture for policymakers.
The shadowstats website has continued with the older more accurate methodology where it shows a more realistic view, when it can.
US data for unemployment has been fundamentally flawed for decades. In essence, the methodology stops counting people looking for work when they have been unable to find work after some amount of time. If you aren't actively looking for work or recently had work, you aren't counted. Actively looking doesn't mean interviews, it means being registered in a work program, which has a large number of requirements and competition.
The resolution foundation has several good analysis of multiple countries that show this trend.
If you want more specific details than that, I'd suggest "The Pinch" by Willett. It has many credible references, supporting this at the back (too many to list).
Where other similarly themed books like "The Theft of a Decade", offer no more than opinion, this covers what you are looking for with gratuitous references.
There is nothing I can do about that.
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.
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.
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.
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.
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.
What you and a large majority of everyone else don't get is this,
AI does not need to become AGI to destroy us all. The minimal much easier to reach requirement it needs to do that is make it so people can't find jobs.
People will do the rest in accordance with fundamental societal mechanics that have been known for centuries.
Economies work in a loop. You trade time for production (labor) in the factor markets, and you store it in money, you use money to buy what you need (food).
Companies buy that labor to produce goods which they then go and sell to fulfill those basic needs.
What happens when basic needs cannot be met, because companies don't buy the labor, and government has a history over the past generation (which is commonly 20 years) of doing largely nothing.
When the people starve, there will be blood in the streets, and there are a lot more of them than there are people at the ideal corporation which would be 1 person at the top with robot slaves replacing everyone else.
The only solution would be a central structure that provided for basic needs without a market. This has been tried many times over the past 100 years, all of them failed (afaik). You run into what's known as the economic calculation problem (or debate) depending on where you look, and a whole host of other difficulties which have no solution.
It works for a time, then something unpredictable happens which causes shortages, then those shortages causes death on a massive scale. There's a lot of written history showing this.