A new paper is out today that tries to estimate job losses from ChatGPT and other similar apps based on Large Language Models. The paper includes much talk of binning and rubrics and OLS regressions, all with lots of Greek letters scattered around. But you don't care about that, since it all boils down to "this is a rough guess."
What you care about is the final result. Which jobs are safe and which are going to disappear thanks to GPT software? The authors present estimates from human experts and from a machine algorithm, and naturally we want to see what the machine says. Here it is:
Oops! You poor humans can't read that, can you? Fine. Here's a slightly more human readable version, with the most vulnerable jobs at the top:
Social workers finally get their due! They are apparently the workers who are safest from computer overlordship in our brave new GPT world. More generally, if your job requires you to work with either people or stuff, you're in decent shape. But if your job primarily requires you to shuffle words and numbers around on a computer, you may want to think about changing careers over the next decade or so.
Whew. Education is safe for humans.
However, given the available occupations, why bother?
Google paid 99 dollars an hour on the internet. Everything I did was basic Οnline w0rk from comfort at hΟme for 5-7 hours per day that I g0t from this office I f0und over the web and they paid me 100 dollars each hour. For more details
visit this article... https://createmaxwealth.blogspot.com
Eve, your job could definitely be done by a computer, I'm sorry to say.
It is being done by a computer.
Is this list taking into account combining it with increasing advances in robotics and remotely controlled devices?
I've often thought something like mining is an obvious candidate for robots since they won't require oxygen or ventilation.
It's proven really challenging to automate non-repetitive manual tasks taking place in environments constructed for human beings.
The environments don't have to be constructed for human beings.
The key here may be removing the "environments constructed for human beings" requirement.
I'd think that in cases like mining the job could be structured in ways that favor automation. It's not like robots would need to handle mining equipment designed for humans -- they will be the mining equipment. For example, they may be able to work in tunnels that are carved to require few or no supports, where a human wouldn't fit.
There are other blue collar jobs, such as construction, where the tasks and even the structures themselves could be redesigned for robot assembly.
Obviously, where the tools of the trade require no physical adaptation and already involve some automation it is much simpler, which is why white collar jobs will go first. But in some cases (for example, automobile assembly) blue collar jobs are already largely taken over by robotics. And advanced in AI for both vision and decision-making coupled with the advances already made in robotic manipulation will help eliminate the rest.
"I'd think that in cases like mining the job could be structured in ways that favor automation."
Automation is already *heavily* favored in mining. Look up what longwall coal mining is, for example. We don't mine things via guys with pickaxes anymore, and automation isn't going to look like an android with a hardhat.
It looks like a common thread with those is that AI is going impact the white-collar workplace faster than robotics can impact the blue collar one. Not sure about that, but it has proven noticeably much harder to automate non-repetitive physical and manual tasks than we thought. There's less incentive to pay for it, either - blue collar labor is pretty cheap in the scheme of things.
Hats off to Paul Krugman for potentially calling that 25 years ago with his "White Collars Turn Blue" essay.
Still not buying that it actually threatens jobs. It will just make them different. Humans will still be required in these jobs or jobs analogous to them.
Also, "The authors present estimates from...[and] a machine algorithm." Lol.
Stable Diffusion is already putting commercial illustrators out of work. Right now. Collateral creation that would previously have gone to working artists is instead being generated.
If your response is that's just a tiny corner of the work force, you're not thinking this through.
I think there's a strong case to be made that LLMs and the like won't adversely affect the overall level of employment, ie, we won't suffer mass unemployment on net.
But individual sectors? Why would the immediate future be any different from past eras? Technology eviscerates lots of jobs. Full stop. How many people are employed as travel agents these days? How about blacksmiths?
I'd read this a little differently -- if your job involves actual knowledge with real world cognitive referents attached, you're safe. But if your job is mechanical, repetitive, or involves pushing numbers according to strict algorithms without regard to their meaning, watch out.
Of course, that way of putting it does not support the "AI monsters under the bed" point of view.
I have little doubt that ChatGPT will rise to the pinnacle of this:
https://philosophynow.org/issues/53/On_Bullshit_by_Harry_Frankfurt
This is the article that prompted me (a few minutes ago) to think of good ol' Harry and bullshit:
https://www.salon.com/2023/03/19/dont-believe-the-hype-why-chatgpt-is-not-the-holy-grail-of-ai-research/
Something relevant from Salon: That's a blast from the past! But yeah, what it said. I am particularly bemused by Microsoft's decision to hand search over to a serial bullshitter. The various pieces swooning over it notably restrict themselves to mushy stuff: give me a recipe for [x], lets have an undergrad philosophy bull session, and the like. They don't ask it a question where the answer is either right, wrong, or too incomplete to be useful. It is a complete disaster. I particularly like how it has no data post-2021. For chat, this is a quirk. For search, it is an absurdity.
I look forward to AI writing trumps social media posts. It can surely do better than “horse face” in describing that lady he had sex with.
I note that lawyers/legal profession isn't even on the list. Maybe because nobody else cares if we lose our jobs? ¯\_(ツ)_/¯
More seriously, I think lawyering is one of those professions that won't be eliminated because of LLM, but will drastically change. The human side of the job is arguably the hardest part to replace, the arguments in court, client management, "know your judge" etc. are likely to be around for a long time. But in tax law, Blue J, a Toronto company, has been providing software services for more than 5 years that can provide legal research and opinion that had a 95% accuracy rate a few years ago, and is probably better now. It's only a matter of time before other companies start training software on judgments in other areas of law, producing similar results.
As noted above, this isn't going to eliminate the need for litigators to go to court or lawyers to explain to their clients what those opinions mean in layperson's terms, but it is going to basically eliminate the need for legal researchers. Legal research is a lengthy process, and one that hitherto has been very time-intensive. There is a powerful economic incentive on clients, personal and corporate, to streamline the expense of legal research and opinion writing. When LLMs are able to take over legal research, a lot of lawyers are going to find themselves without jobs.
Of course it will have to break its habit of making shit up. I wouldn't think this is an unsolvable problem, but I haven't seen articles by knowing people explaining how it will be fixed.
Just a half hour ago I saw someone had asked all the big chat programs the question "Which is heavier, 1 kilogram of steel or 2 kilograms of feathers?" and every single one spat out a long, flowery answer about how they both weight exactly the same. XD
I know Kevin and other folks are bullish on these chatbots being able to do everything within a couple years but I just don't see it. Among many other issues, there's no way to weed out the wrong answers or made-up information (short of fact-checking the entire set of source data, which is impossible), and if there's no guarantee of perfect accuracy, I don't see what the point is beyond a cute party trick. (And, of course, disrupting entire fields and putting people out of work because a bunch of people like Kevin and some tech billionaires are pushing the idea that chatbots can do what they can't actually do.)
Travel agents, and before them the office secretarial pool, and before them the legions of switchboard operators, and before them the vast infrastructure that supported a transportation industry dependent on horses, wagons, and carriages....
See a pattern? New technology upends existing industries, but they morph and adapt. The telephone "industry" seems pretty robust. Every mall in North America is bristling with storefronts dedicated to cell phones, accessories, and services. Office work has changed a lot since Mad Men, but the sector seems to have survived okay. A whole lot of travel agents lost their jobs, but the travel and tourism industry seems to be ticking along (following some recent hiccups due to a global health emergency).