As with ever other major innovation, it will both destroy and create jobs. Whether AI creates more good than harm will be hard to actually evaluate. Like every innovation, there will be general disagreement on what constitutes "good".
> As with ever other major innovation, it will both destroy and create jobs.
Yep.
Major innovations change economic equilibriums and as the economy shifts accordingly winners and losers are produced. Economic losers who see what's happening get angry and oppose the innovation. In extreme cases this leads to social problems.
IMO a key role of government is shepharding societies through major economic shifts while minimising social instability. IMO it's more productive to accept change and plan for it than try to prevent it.
> Whether AI creates more good than harm will be hard to actually evaluate.
Perhaps I'm excessively fatalist, but I think that's largely irrelevant.
Even if we had an oracle which could tell us that a given innovation will be a net negative in the long term, any innovation still presents opportunities for individuals and groups to gain advantages over others in the short to medium term.
WMD proliferation is the only example I can think of of a collective decision to slow development of a particular technology due to a belief it will ultimately be a net negative. Even then we haven't stopped, we only slowed after major players consolidated their positions, and we've yet to see how successful that slowing will ultimately be. As in - did we put the cat back in the bag permanently, or will it get out again?
Climate action might be heading that way, too, but I don't think the cat's even back in the bag yet.
The problem is, it doesn't take much destroying. Hitler rose to power off the back of 3 years of 25% unemployment.[0] It just doesn't need to replace all jobs, nor does it need to replace them permanently to have drastic, world changing repercussions.
My take is that the real societal shift will come when we have AI-powered bipedal robots, a la Agility Robotics' Digit or Tesla's Optimus, at something like <$100,000 a pop.
AI in knowledge work will probably see the sort of changes people talk about - jobs destroyed but new jobs created. A robot that can do cleaning, construction, even retail sales - that's going to wipe out swathes of jobs. You'll have armies of workers that can work nonstop, don't steal, don't want raises, don't have a bad day at work, and can be retrained as quickly as you can install new software.
I don't see that there's any reason we won't be able to have those kinds of robots in, conservatively, 20 years, and I don't see how we get close to replacing the number of jobs they take.
Yeah, that's totally fair. I suppose I see it as different than human problems for two reasons.
First, workers become fungible - right now if John who's the expert on the particular machine we need to use today just got dumped and is doing a bad job, we have to either work with him in his downtrodden state or find someone else with his expertise. With robots, when one starts malfunctioning you swap it out.
Second, you can diagnose and fix it. I can't fix my employee's broken heart or lack of sleep because they have a small child at home. Whatever is wrong with the robot, I can fix.
I cannot wrap my head around how far awry society has gone, that "eliminating tedious labor that no one really wants to do" gets reframed as "destroying jobs". Obviously in reality the benefits do not immediately flow to the people relieved of work, yet this obvious critical flaw of capitalism is always blamed on the automation in question. Yet few suggest we destroy our existing automation, no matter how many wonderful jobs it would recreate such as "washerwoman" and "water carrier". Society is clearly wealthier, the more "jobs" we eliminate.
That’s just your opinion on what jobs are “fun” and “not fun”. The reason those people do those jobs isn’t because it fulfills their 7th level of Maslow’s hierarchy, it’s because it puts foods on the table.
So when that’s taken away from them, what’s left? Join a coding bootcamp?
Having an abundance of resources and people having nothing to do seems like an easier problem than lack of resources and people working 12 hours a day.
> Having an abundance of resources and people having nothing to do
Many revolutions have started under exactly these conditions. The problem is that some person or group is going to end up owning those resources, preventing their benefits from flowing to the people "with nothing to do". It's not a perfect dichotomy of either work 12 hours a day or do nothing, some balance has to be struck such that everyone can do something to contribute to society in a meaningful way.
I once dreamed of a day when robotics would advance to a point that humans would be free of labor. What would they do with all of their newfound time? I imagined us writing, painting, philosophising, exploring the depths of arts, sciences mathematics. In a darker side of the dream, humans are left without purpose, and chase dopamine instead: the species dies with a heroin-muffled whimper.
Stable diffusion poured cold water on that dream. We merely feed the computer all digitized creative works, et voilá, it's more creative than 99% of humanity, and more productive to boot. And it turns out that plumbing is really hard. Therapy, not so much.
With AI poised to replace us in knowledge work, we are left with two options: hard labor, or drugs. And a sufficiently advanced AI could pretty quickly make a plumber-bot.
The only obstacle to this nightmare future is that capitalists are winning the AI game. For the foreseeable future, robots will exist to serve the capitalists. If nobody can afford to live in robo-utopia, it collapses under its own expense.
20 years is not conservative; it is a very aggressive timetable for significant impact on employment.
I think you underestimate just what is involved in most "unskilled", manual jobs, and you also underestimate scale-up difficulties in production and teaching the robots. These are not a bunch of rectangular slabs with only one moving part.
Further, there are hundreds of thousands of occupations; the most numerous occupation, health care aide, is extremely diverse. It's not like "write the next iOS or Android, and you're done": you have to repeat that on the order of a million times.
(Note: I am not given to using adverbs, especially intensifiers, as a rule. Please take that into account when reading "very" and "extremely" above.)
Counterpoint, 20 years ago smartphones didn't exist. Now it's been nearly impossible to work without one for around a decade. Technology can move quite fast sometimes.
> Now it's been nearly impossible to work without one for around a decade.
How true this is probably depends on where you live. In my part of the US, this isn't even close to being true. You can get along just fine without a smartphone.
Where I live, in the UK, it isn't possible to login to three of my bank accounts for admin or make credit card payments without a smartphone to confirm payments sometimes. I had to approve the debit card purchase of a small pizza via in-app notification from the bank last week.
Email, voice phone and SMS authentication used to be used but they are gradually being deprecated by some financial services.
One job ago I had to use my smartphone several times a day to approve corporate VPN logins on the work-provided computer, before I could access the Git repo, BitBucket and Linux dev VMs. All of us had to install a specific authenticator app on our phone.
Two jobs ago I had to use Google auth with my smartphone used as a security token, to access work documents and emails. When the phone broke I lost that access, despite transferring a backup to a new phone and most things surviving the transfer, because that access was associated to the device.
I have a strict policy about work and smartphones. Nothing related to work touches my smartphone. If an employer has processes that require a smartphone, then the employer needs to supply that smartphone to me, just like my employer needs to supply other essential work equipment such as a computer.
The brightest side of AI in my opinion is that the jobs created will be of _lower_ technical skill compared to the jobs created by the computer revolution.
It will be _easier_ to take advantage of AI than it was to take advantage of a computer.
This means the new jobs could potentially be ones that more people can perform and the jobs eliminated will be some of the highly-technical ones.
The irony is my company is starting it's annual hackathon week. The company has gone out of its way to make Google AI available for us, met with Google employees who are rolling out the red carpet to give us access to unreleased tools, features and experimental apps. We were already tightly partnered with google before this, so it all makes sense google wants to see we very successful, meaning we can never leave them for AWS or Azure.
I feel like each team is looking for a way to optimize another teams work, making them more expendable after we fully adopt AI systems. Really feels like a race to the bottom, luckily only a handful of people know what I do and they don't have access to my systems to even start trying to figure out how I can be replaced. That's called job security my friends...
I believe anyone saying AI isn't going to reduce jobs is full of crap, it's going to happen.
I already see ai replacing jobs of people who tend to handle communication where a friend of mine is doing the work of 5 by generating 90 percent of their comms and hand modifying the output.
Basically if your job involves finding a way to tell people information you are replaceable by current ai models. If your job requires creative thinking, you're safe, ai ain't even close to that.
Literary translation is a field where machines have already threatened creative thinking jobs. Already even some respectable publishers in various countries are machine-translating novels, giving them a light copyediting (not always a deep editing -- the industry has been cutting back on that for years) by someone who doesn't speak the source language, and selling them to the public. It is not that machine translation can already think so creatively, it is just that the cost savings are so great that human creativity is viewed as an unaffordable luxury and the market tolerates the quality hit.
I agree here. This is always the way. I mean literally most companies I work for are all about "how do we take a human process and make it automated, and thus more accurate and better!"
26 comments
[ 3.0 ms ] story [ 61.9 ms ] threadYep.
Major innovations change economic equilibriums and as the economy shifts accordingly winners and losers are produced. Economic losers who see what's happening get angry and oppose the innovation. In extreme cases this leads to social problems.
IMO a key role of government is shepharding societies through major economic shifts while minimising social instability. IMO it's more productive to accept change and plan for it than try to prevent it.
> Whether AI creates more good than harm will be hard to actually evaluate.
Perhaps I'm excessively fatalist, but I think that's largely irrelevant.
Even if we had an oracle which could tell us that a given innovation will be a net negative in the long term, any innovation still presents opportunities for individuals and groups to gain advantages over others in the short to medium term.
WMD proliferation is the only example I can think of of a collective decision to slow development of a particular technology due to a belief it will ultimately be a net negative. Even then we haven't stopped, we only slowed after major players consolidated their positions, and we've yet to see how successful that slowing will ultimately be. As in - did we put the cat back in the bag permanently, or will it get out again?
Climate action might be heading that way, too, but I don't think the cat's even back in the bag yet.
[0]Fantastic discussion with Suleyman and Yuval Harari https://youtu.be/7JkPWHr7sTY
AI in knowledge work will probably see the sort of changes people talk about - jobs destroyed but new jobs created. A robot that can do cleaning, construction, even retail sales - that's going to wipe out swathes of jobs. You'll have armies of workers that can work nonstop, don't steal, don't want raises, don't have a bad day at work, and can be retrained as quickly as you can install new software.
I don't see that there's any reason we won't be able to have those kinds of robots in, conservatively, 20 years, and I don't see how we get close to replacing the number of jobs they take.
All systems break. Complex systems break in complex ways. Robots will definitely have bad days.
First, workers become fungible - right now if John who's the expert on the particular machine we need to use today just got dumped and is doing a bad job, we have to either work with him in his downtrodden state or find someone else with his expertise. With robots, when one starts malfunctioning you swap it out.
Second, you can diagnose and fix it. I can't fix my employee's broken heart or lack of sleep because they have a small child at home. Whatever is wrong with the robot, I can fix.
So when that’s taken away from them, what’s left? Join a coding bootcamp?
Many revolutions have started under exactly these conditions. The problem is that some person or group is going to end up owning those resources, preventing their benefits from flowing to the people "with nothing to do". It's not a perfect dichotomy of either work 12 hours a day or do nothing, some balance has to be struck such that everyone can do something to contribute to society in a meaningful way.
Stable diffusion poured cold water on that dream. We merely feed the computer all digitized creative works, et voilá, it's more creative than 99% of humanity, and more productive to boot. And it turns out that plumbing is really hard. Therapy, not so much.
With AI poised to replace us in knowledge work, we are left with two options: hard labor, or drugs. And a sufficiently advanced AI could pretty quickly make a plumber-bot.
The only obstacle to this nightmare future is that capitalists are winning the AI game. For the foreseeable future, robots will exist to serve the capitalists. If nobody can afford to live in robo-utopia, it collapses under its own expense.
I think you underestimate just what is involved in most "unskilled", manual jobs, and you also underestimate scale-up difficulties in production and teaching the robots. These are not a bunch of rectangular slabs with only one moving part.
Further, there are hundreds of thousands of occupations; the most numerous occupation, health care aide, is extremely diverse. It's not like "write the next iOS or Android, and you're done": you have to repeat that on the order of a million times.
(Note: I am not given to using adverbs, especially intensifiers, as a rule. Please take that into account when reading "very" and "extremely" above.)
I think humanoid AI robots are a bigger leap.
How true this is probably depends on where you live. In my part of the US, this isn't even close to being true. You can get along just fine without a smartphone.
Email, voice phone and SMS authentication used to be used but they are gradually being deprecated by some financial services.
One job ago I had to use my smartphone several times a day to approve corporate VPN logins on the work-provided computer, before I could access the Git repo, BitBucket and Linux dev VMs. All of us had to install a specific authenticator app on our phone.
Two jobs ago I had to use Google auth with my smartphone used as a security token, to access work documents and emails. When the phone broke I lost that access, despite transferring a backup to a new phone and most things surviving the transfer, because that access was associated to the device.
It will be _easier_ to take advantage of AI than it was to take advantage of a computer.
This means the new jobs could potentially be ones that more people can perform and the jobs eliminated will be some of the highly-technical ones.
I feel like each team is looking for a way to optimize another teams work, making them more expendable after we fully adopt AI systems. Really feels like a race to the bottom, luckily only a handful of people know what I do and they don't have access to my systems to even start trying to figure out how I can be replaced. That's called job security my friends...
I believe anyone saying AI isn't going to reduce jobs is full of crap, it's going to happen.
Basically if your job involves finding a way to tell people information you are replaceable by current ai models. If your job requires creative thinking, you're safe, ai ain't even close to that.