GPT-4 will be an improvement for most AGI cultists, but also a huge disappointment. The models from other companies that follow will only allow us to improve most professionals, not replace them.
AI Winter is coming. It will start slowly, but we have just reached the peak. The next five years will be a journey of discovering the limits of the technology and killing the dumb hype created by VCs and Musk boys.
AI winter is coming no matter what, the hardware for any sort of AGI doesn't exist and wont exist for the forseeable future
GPT-4 will be an improvement for most AGI cultists, but also a huge disappointment. The models from other companies that follow will only allow us to improve most professionals, not replace them.
AI Winter is coming. It will start slowly, but we have just reached the peak. The next five years will be a journey of discovering the limits of the technology and killing the dumb hype created by VCs and Musk boys.
This is likely, IRL we don't see catastrophic asymptotic behavior very often, no matter how much our pattern matching brains want it to happen.
But we know AGI is possible, (being NGIs) just not how close we are. So I won't be surprised if the AI econo-rapture happens next year either.
>improve most professionals, not replace them
What this means is replace 80% of professionals with the other 20% + AI
It won't destroy 100% of all professions, just probably destroy YOUR job
>GPT-4 will be an improvement for most AGI cultists, but also a huge disappointment.
What people don't seem to understand is that these language models are only engineered to produce realistic text. There is nothing in the transformer architecture that validates the logic or facts stated in the produced text.
There are attempts to add this feature to generative language models, but I don't know the state of those.
>The next five years will be a journey of discovering the limits of the technology and killing the dumb hype created by VCs and Musk boys.
Probably. The good thing is that language models have made empty words almost free to produce, so the entire copywriter industry ought to die soon.
I work in army
Battle robots are not a thing until we come up with something like artificial muscles from mechwarrior, though drones have already altered modern doctrines
I do computer vision/hpc stuff.
My job will get replaced eventually, I don't expect us to stay on our current trajectory though, eventually we'll reach a plateau, where we can't realistically scale up anymore due to prohibitive costs and then AI will stagnate until we get a new fundamental breakthrough.
This. Tbh I believe that groundbreaking stuff will come from people with less resources. People doing stuff on smaller servers, personal computers, something like that. Something that works insanely well on shitty hardware and presents god-like behavior when scaled to what is being use today to create dumb responses that only impress normies (ChatGPT).
I think the next breakthrough is going to be something like letting the model make requests to an external database (or even query google) to facilitate its responses.
Right now we have a good chunk of the internet encoded purely in the model's weights, that's obviously going to be insanely resource intensive.
I'm not sure what architecture would work best for this kind of thing, but there's plenty of room for experimentation.
This. I'd love it if a machine could take my job as a web developer. Then I could use that same tech to make my own shit lightning fast. But until then, I'll continue earning a crazy high salary and retain excellent work-life balance.
Imagine being some cuck retard wearing a gay suit all day to suck corporate cock to hope for a few extra bucks.
AI will replace programmers last. Quote me
I do nothing for a living, a computer cant do that, even if it's doing nothing, it still has to operate the NOP command.so in doing nothing i am more advanced than AI
https://en.wikipedia.org/wiki/NOP_(code)
Even when idling, your body is doing some incredibly complex stuff. If we had a machine that could emulate the liver, we could save so many people. Now only chinese political captives are our only hope
Im a "pc operator", i write little scripts, use sql, excel, php, some corporate crap and i translate retarded gibberish of my colleagues into human language for full time programmer that suffers writing yii code.
My work won't be replaced because i live in a third world country and 95% of the population are cretins who won't be able to express their request to ai even if their life depended in it.
>cretins who won't be able to express their request to ai even if their life depended in it.
Isn't strength of the AI _exactly_ that it is able to understand such retards, instead of needing expensive middlemen (aka programmers)?
No, because a big part of understanding brainlet creetins is to communicate to them so you get a proper output.
The person you are quoting real job isn't writing the code, its playing advanced pokeface reading when repeating requests and guessing where the gigantic holes in their vocabulary are located.
What the AI will do is give faster output, the person will spot its wrong, and then get confused without realizing they are in error.
It won't fix that the user is lacking vocabulary, or are unfamiliar with concepts.
Medical doctor. While my knowledge can probably be replaced by AI already, there are some manual things I have to do. Robotics isn't advanced enough for that yet.
People also tend to trust a person more than a machine.
>People also tend to trust a person more than a machine.
This is why I think doctors won't be replaced any time soon, for better or for worse. There are people who do dumb shit because they can't even trust doctors, and there are people that choose their doctors not because of expertise but because they give off a "better impression".
Hardware technician. Machine learning could replace my job, but I doubt it would ever be cost effective. Designing a robot to plug cables in would cost more than I'll ever make in a lifetime.
I work for the government. Sysadmin
1. It'll be years to approve an AI, so I have a bit of a window to transition jobs
2. Grifters get their way in gov and the extent people will go to to grift will prevent ai from being a thing.
3. We can't automate our imaging process as-is, let alone have an AI do it.
4. The ai would need to also be racist, which it inherently is, but God knows it'll be nerfed before it touches gov sysadmin
As an engineer who builds and maintains bespoke complex machines, I think my job will last longer than most. AI might take over design work but going out, analysing and fixing mechanical/hydraulic/pneumatic/electrical problems with a large complex finely tuned machine is beyond the horizon of current ai/robotics.
I'm a nurse. job still requires a meatbag to do the wagie stuff and i'm not sure that robotics will become that advanced in a short 30 years. maybe, hopefully, but i doubt it. Don't get me wrong, i don't think my job is anything special, we're more or less glorified tradies, but some of those jobs seem they would be trickier to replace (just longer time i suppose)
accountant. i'm the fall guy for the CFO, who would otherwise go to jail for materially misleading or fraudulent financial statements following the sarbanes-oxley act.
>Tell me your job and explain to me why machine learning won't replace it within thirty years.
I'm a truck driver. People have been predicting self driving cars replacing us for over 40 years now. It never happens. Nothing ever happens.
Logistics Coordinator, i clean up human made errors, until everything upstream of me is automated, a computer couldn’t realistically replace me.
“Why won’t this pallet scan” >the supplier didn’t even attach the labels but just wedged them inbetween the boxes >or it has the entirely wrong labels for the pallet >they hand wrote on a destination location and nothing more than that >it’s actually an empty pallet of returnable containers they marked with a label but for zero pieces >the pallet is marked free movement and has no digital signature, only label is an address
Given enough time I’m positive you can train a computer to deal with this nonsense, but it would be much easier for said computer if that nonsense didn’t happen. Trying to understand why humans do stupid shit isn’t what computers are good at.
>It's a meme that requires exponentially more resources to be remotely useful
that's what they said about computers
imagine someone claiming "feature films will be rendered on these one day", looking at a '70s computer spending minutes drawing a wireframe house
"oh they'd have to get millions of times faster for that to be a concern"
well now they're millions of times faster, and features films are being rendered on them
Electrical Engineering. I draw schematics for industry machines. But i mostly just copy and paste from the old ones, since the machines are almost the same each time anyways. I could see getting replaced by some machine tbh, i would just change into a field that isn't automatable (yet)
I work with angry old people and replace their keyboards and install Adobe Reader. The last thing they're going to trust is an AI. Then again, in 30 years I'll probably be one of the angry old people...
no stinking machine is ever be as good as i am at getting drunk, slacking off, collecting welfare, running scams and being the best damn public nuisance you've ever seen.
>tire technician
You already have machines that can automatically de-bead, replace the pneumatic valves, and balance them. For common types of automobile tires.
So long it remains a shit underpaid job, and the most complex part of the job is the jacking process and edge cases... automated technique will remain non competitive.
For agriculture equipment beading and de beading large tractor wheels is complicated, since most shops lack proper routines for storage, and they are large and heavy enough to deform into a fucking mess when improperly stored or transported.
Storage and logistics could be automated away, but its not happened yet. People are content with manually indexing with excel or whatever spahgetti the CRM deals with.
I would like to add that for truck wheels there already exist automated wheel unbolter, both pneumatic and electric. There is a lot of fancy automation that has already happened. The balancing machines are already 50% automated as well. I forsee more in the future, but a lot of the low hanging fruits has happened.
Do I forsee my job going away?
No, I will leave it once I get my education/CV back in shape.
But if you managed to create a completely automated system that include jacking and wheel change, there could be money in that. But its currently not a significant cost, when the rims and rubber is the core cost, and even understaffed shops has no problem taking on the fly jobs. Its also a job that exist because people want their undercarriage, suspension, rims and brakes inspected by a unskilled operator, as renting a proper room sized jack seem to be a meme.
I maintain ancient humongous backend systems for retail. It's so big that the cost of automating it is so enormous that it's probably never gonna happen. (As long as we keep it running)
Oh they've tried though, have lost some billions here and there over the years. =)
I post racist and antisemitic rants on anonymous forums for $0 per hour, it will never be replaced because the technology will be gelded to be inclusive and respectful to minorities as soon as it hits mainstream.
I think it's theoretically possible to automate the world in the same way it's theoretically possible to operate at peak resource management and live in prosperity for all mankind
Over the next few decades you're going to see companies waste billions trying to automate their workforce away with a few "success" stories that end up creating more jobs than we started with. It's going to be the eighties with digitization all over again
>It's going to be the eighties with digitization all over again
Yes.
But some parts of the world have now deindustrialized. So without access to primary and secondary resources... can you truly grow?
i take things that aren't digital (photos, film, tapes (ok some of these are digital)) and i make them digital
ai won't take my job because 99% of my job is being a real person talking to real old people who don't like machines and just want to see their grandfathers from 1912 again
the other 1% is putting the thing into the other thing and pressing the button, which i'm sure AI could do but it won't because noone would pay a computer to help them use or understand a computer
Carpenter. Everything robots can do in my job is already done by robots. For custom and small series productions, robots are far too expensive to program and maintain. If an AI could do all my work, there's still the need to buy expensive and unreliable bots to get to the site and actually do the job, and they need to be strong enough to move modules around.
>thirty years
I'll probably be dead zoom zoom
Iam 24 but I probably not going to make it another 30 years on this shithole planet especially the way things are goning
in thirty years my country is a fucking warzone
realistically, these
AI winter is coming no matter what, the hardware for any sort of AGI doesn't exist and wont exist for the forseeable future
GPT-4 will be an improvement for most AGI cultists, but also a huge disappointment. The models from other companies that follow will only allow us to improve most professionals, not replace them.
AI Winter is coming. It will start slowly, but we have just reached the peak. The next five years will be a journey of discovering the limits of the technology and killing the dumb hype created by VCs and Musk boys.
>Machine learning will make people's lives easier rather than hellish
NNNOOOOOOOOOO NNOOOO
This is likely, IRL we don't see catastrophic asymptotic behavior very often, no matter how much our pattern matching brains want it to happen.
But we know AGI is possible, (being NGIs) just not how close we are. So I won't be surprised if the AI econo-rapture happens next year either.
>but we have just reached the peak
this is just peak oil nonsense all over again
U wot m8
>improve most professionals, not replace them
What this means is replace 80% of professionals with the other 20% + AI
It won't destroy 100% of all professions, just probably destroy YOUR job
>2020
This is the peak bro, AI isn't shit
>2021
This is the peak bro, AI isn't shit
>2022
This is the peak bro, AI isn't shit
>GPT-4 will be an improvement for most AGI cultists, but also a huge disappointment.
What people don't seem to understand is that these language models are only engineered to produce realistic text. There is nothing in the transformer architecture that validates the logic or facts stated in the produced text.
There are attempts to add this feature to generative language models, but I don't know the state of those.
>The next five years will be a journey of discovering the limits of the technology and killing the dumb hype created by VCs and Musk boys.
Probably. The good thing is that language models have made empty words almost free to produce, so the entire copywriter industry ought to die soon.
Well I'm currently an IT advisor so my job is telling people how to replace you, if anything.
My job is being Mommy's Special Boy and no robot could ever replace me
I work in army
Battle robots are not a thing until we come up with something like artificial muscles from mechwarrior, though drones have already altered modern doctrines
I do computer vision/hpc stuff.
My job will get replaced eventually, I don't expect us to stay on our current trajectory though, eventually we'll reach a plateau, where we can't realistically scale up anymore due to prohibitive costs and then AI will stagnate until we get a new fundamental breakthrough.
This. Tbh I believe that groundbreaking stuff will come from people with less resources. People doing stuff on smaller servers, personal computers, something like that. Something that works insanely well on shitty hardware and presents god-like behavior when scaled to what is being use today to create dumb responses that only impress normies (ChatGPT).
I think the next breakthrough is going to be something like letting the model make requests to an external database (or even query google) to facilitate its responses.
Right now we have a good chunk of the internet encoded purely in the model's weights, that's obviously going to be insanely resource intensive.
I'm not sure what architecture would work best for this kind of thing, but there's plenty of room for experimentation.
a.i. doesn't eat tendies.
I'm a software developer. I make a web app do what the customer wants
>why machine learning won't replace it within thirty years.
The customer doesn't know what they want.
This. I'd love it if a machine could take my job as a web developer. Then I could use that same tech to make my own shit lightning fast. But until then, I'll continue earning a crazy high salary and retain excellent work-life balance.
Imagine being some cuck retard wearing a gay suit all day to suck corporate cock to hope for a few extra bucks.
AI will replace programmers last. Quote me
You do realize that 30 years is forever in current technological era? 1984 was thirty years ago
I do nothing for a living, a computer cant do that, even if it's doing nothing, it still has to operate the NOP command.so in doing nothing i am more advanced than AI
https://en.wikipedia.org/wiki/NOP_(code)
Even when idling, your body is doing some incredibly complex stuff. If we had a machine that could emulate the liver, we could save so many people. Now only chinese political captives are our only hope
>If we had a machine that could emulate the liver, we could save so many people
imagine if we did that instead of chatGPT
Senior hand on a lettuce farm.
Because my job is actually important.
Im a "pc operator", i write little scripts, use sql, excel, php, some corporate crap and i translate retarded gibberish of my colleagues into human language for full time programmer that suffers writing yii code.
My work won't be replaced because i live in a third world country and 95% of the population are cretins who won't be able to express their request to ai even if their life depended in it.
>cretins who won't be able to express their request to ai even if their life depended in it.
Isn't strength of the AI _exactly_ that it is able to understand such retards, instead of needing expensive middlemen (aka programmers)?
When was the last time you watched a normie interact with an IVR ? It would be many times worse.
No, because a big part of understanding brainlet creetins is to communicate to them so you get a proper output.
The person you are quoting real job isn't writing the code, its playing advanced pokeface reading when repeating requests and guessing where the gigantic holes in their vocabulary are located.
What the AI will do is give faster output, the person will spot its wrong, and then get confused without realizing they are in error.
It won't fix that the user is lacking vocabulary, or are unfamiliar with concepts.
3D modeling
God I hope it replaces me way sooner than 30 years.
I'm American politician.
We could replace you with a random number generator and it would improve performance.
I shall made it illegal
>Flips a coin
This coin?
https://www.wsj.com/articles/janet-yellen-dismisses-minting-1-trillion-coin-to-avoid-default-11674417541
real?
I sell things and people will always want things
Robots are always limited because humans made them
>math teacher
>It took me half an hour to make chatgpt understand that R is not an infinite-dimensional vector space
Medical doctor. While my knowledge can probably be replaced by AI already, there are some manual things I have to do. Robotics isn't advanced enough for that yet.
People also tend to trust a person more than a machine.
>People also tend to trust a person more than a machine.
This is why I think doctors won't be replaced any time soon, for better or for worse. There are people who do dumb shit because they can't even trust doctors, and there are people that choose their doctors not because of expertise but because they give off a "better impression".
>JUST IN - Microsoft to invest multi-billion dollars in ChatGPT creator OpenAI days after cutting 10,000 human jobs.
Hardware technician. Machine learning could replace my job, but I doubt it would ever be cost effective. Designing a robot to plug cables in would cost more than I'll ever make in a lifetime.
Technician, railway signalling.
We're unionized.
I work on developing AI and other ML stuff
If AI replaces me we get the singularity, after that nothing really matters anymore
my job is picrel
if machine learning replaces it I'll just kill myself
I work for the government. Sysadmin
1. It'll be years to approve an AI, so I have a bit of a window to transition jobs
2. Grifters get their way in gov and the extent people will go to to grift will prevent ai from being a thing.
3. We can't automate our imaging process as-is, let alone have an AI do it.
4. The ai would need to also be racist, which it inherently is, but God knows it'll be nerfed before it touches gov sysadmin
>which it inherently is
that's because the people who built it have racial biases. AI isn't inherently anything, that's 80 IQ luddite thinking
reality has a racist bias, sorry
As an engineer who builds and maintains bespoke complex machines, I think my job will last longer than most. AI might take over design work but going out, analysing and fixing mechanical/hydraulic/pneumatic/electrical problems with a large complex finely tuned machine is beyond the horizon of current ai/robotics.
I'm a nurse. job still requires a meatbag to do the wagie stuff and i'm not sure that robotics will become that advanced in a short 30 years. maybe, hopefully, but i doubt it. Don't get me wrong, i don't think my job is anything special, we're more or less glorified tradies, but some of those jobs seem they would be trickier to replace (just longer time i suppose)
Professional NEET
You see, the machines will be the ones working for me, and we will be good friends
Shit will get taxed like hell.
Because machine learning will be banned as soon as it threatens the jobs of politicians.
So true.
accountant. i'm the fall guy for the CFO, who would otherwise go to jail for materially misleading or fraudulent financial statements following the sarbanes-oxley act.
I develop AI.
>Tell me your job and explain to me why machine learning won't replace it within thirty years.
I'm a truck driver. People have been predicting self driving cars replacing us for over 40 years now. It never happens. Nothing ever happens.
I'm a student.
Logistics Coordinator, i clean up human made errors, until everything upstream of me is automated, a computer couldn’t realistically replace me.
“Why won’t this pallet scan”
>the supplier didn’t even attach the labels but just wedged them inbetween the boxes
>or it has the entirely wrong labels for the pallet
>they hand wrote on a destination location and nothing more than that
>it’s actually an empty pallet of returnable containers they marked with a label but for zero pieces
>the pallet is marked free movement and has no digital signature, only label is an address
Given enough time I’m positive you can train a computer to deal with this nonsense, but it would be much easier for said computer if that nonsense didn’t happen. Trying to understand why humans do stupid shit isn’t what computers are good at.
>Why is my shipping container being unloaded in a city on the opposite side of the country?
your job sounds pretty safe
right up until all the incoming work is done by robots who don't make these kinds of mistakes in the first place
>why machine learning won't replace it within thirty years.
It's a meme that requires exponentially more resources to be remotely useful
>It's a meme that requires exponentially more resources to be remotely useful
that's what they said about computers
imagine someone claiming "feature films will be rendered on these one day", looking at a '70s computer spending minutes drawing a wireframe house
"oh they'd have to get millions of times faster for that to be a concern"
well now they're millions of times faster, and features films are being rendered on them
because walking around fixing some measurement device and setting up the next experiment is quite hard to automate.
Electrical Engineering. I draw schematics for industry machines. But i mostly just copy and paste from the old ones, since the machines are almost the same each time anyways. I could see getting replaced by some machine tbh, i would just change into a field that isn't automatable (yet)
>teach AI how to use an EDA and copy-paste reference designs and firmwares
>90% of EE jobs vanish
My job will be made illegal within twenty, regardless of the tech industry.
I fuck bitches
Unless my company can afford robots with guns, and I'm pretty sure they can't, Security work is probably safe for awhile.
I'm weighing whether I want to be dead or not before then. On one hand - Sexbots - on the other hand, Robocop.
Not a white-collar job. Simple as. What I do would require AI in android bodies.
>Hehe, while all these morons are getting collared up, I'm designing my own collar!
I’m writing the machine learning for it.
>I’m writing the machine learning for it.
show me the way, anon
I work with angry old people and replace their keyboards and install Adobe Reader. The last thing they're going to trust is an AI. Then again, in 30 years I'll probably be one of the angry old people...
no stinking machine is ever be as good as i am at getting drunk, slacking off, collecting welfare, running scams and being the best damn public nuisance you've ever seen.
I am a fluffer on a porn set
>tire technician
You already have machines that can automatically de-bead, replace the pneumatic valves, and balance them. For common types of automobile tires.
So long it remains a shit underpaid job, and the most complex part of the job is the jacking process and edge cases... automated technique will remain non competitive.
For agriculture equipment beading and de beading large tractor wheels is complicated, since most shops lack proper routines for storage, and they are large and heavy enough to deform into a fucking mess when improperly stored or transported.
Storage and logistics could be automated away, but its not happened yet. People are content with manually indexing with excel or whatever spahgetti the CRM deals with.
I would like to add that for truck wheels there already exist automated wheel unbolter, both pneumatic and electric. There is a lot of fancy automation that has already happened. The balancing machines are already 50% automated as well. I forsee more in the future, but a lot of the low hanging fruits has happened.
Do I forsee my job going away?
No, I will leave it once I get my education/CV back in shape.
But if you managed to create a completely automated system that include jacking and wheel change, there could be money in that. But its currently not a significant cost, when the rims and rubber is the core cost, and even understaffed shops has no problem taking on the fly jobs. Its also a job that exist because people want their undercarriage, suspension, rims and brakes inspected by a unskilled operator, as renting a proper room sized jack seem to be a meme.
no
I maintain ancient humongous backend systems for retail. It's so big that the cost of automating it is so enormous that it's probably never gonna happen. (As long as we keep it running)
Oh they've tried though, have lost some billions here and there over the years. =)
I post racist and antisemitic rants on anonymous forums for $0 per hour, it will never be replaced because the technology will be gelded to be inclusive and respectful to minorities as soon as it hits mainstream.
I think it's theoretically possible to automate the world in the same way it's theoretically possible to operate at peak resource management and live in prosperity for all mankind
Over the next few decades you're going to see companies waste billions trying to automate their workforce away with a few "success" stories that end up creating more jobs than we started with. It's going to be the eighties with digitization all over again
>It's going to be the eighties with digitization all over again
Yes.
But some parts of the world have now deindustrialized. So without access to primary and secondary resources... can you truly grow?
I'll be living in the woods on my homestead by then
i take things that aren't digital (photos, film, tapes (ok some of these are digital)) and i make them digital
ai won't take my job because 99% of my job is being a real person talking to real old people who don't like machines and just want to see their grandfathers from 1912 again
the other 1% is putting the thing into the other thing and pressing the button, which i'm sure AI could do but it won't because noone would pay a computer to help them use or understand a computer
Carpenter. Everything robots can do in my job is already done by robots. For custom and small series productions, robots are far too expensive to program and maintain. If an AI could do all my work, there's still the need to buy expensive and unreliable bots to get to the site and actually do the job, and they need to be strong enough to move modules around.