I’ve been working with gpt-4 since the week it came out, and I guarantee you that even if it never became any more advanced, it could already put at least 30% of the white collar workforce out of business.
The only reason it hasn’t is because companies have barely started to comprehend what it can do.
Within 5 years the entire world will have been revolutionized by this technology. Jobs will evaporate faster than anyone is talking about.
If you’re very smart, and you begin to use gpt-4 to write the tools that will replace you, then you MIGHT have 10 good years left in this economy before humans are all but obsolete.
If you’re not staying up nights, scared shitless by what’s coming, it’s because you don’t really understand what gpt-4 can do.
You sound like one of those idiots preaching the apocalypse from a street corner. Humans obsolete in 10 years? Yeah sure buddy, right after all those profits trickle down. This is just another tool, an interesting one to be sure, but still just a tool. If you’re staying up nights worrying about this, you don’t really understand the technology, or maybe you’re just worried someone is going to realize you don’t do shit.
I work with AI stuff, just getting into LLM, but I have been doing SD work since the public release last year. In just over 1 year the SD capability has gone from being able to draw a passable image of a cat at 512x512 pixels that required a reasonably powerful graphics card to complete to being able to create 4k images on the same cards that are nearly indistinguishable from actual photos/paintings. It is the single fastest adaptation and development of a technology I have seen in my 30 years in tech. I have actually been tracking the job market and the impacts that this will have and he is not all that far off in his estimate. The current push in AI development is nearly a ubiquitous existential threat to employment as we view it in the society of the United States. Everyone is on the chopping block and you’d best believe that the C-level executives want to eliminate as many positions as possible. Labor is viewed as an atrocious expense and the first place that cuts should be made. I challenge you to actually come up with a list of 10 jobs that employ more than 100,000 people in the country that you think would be safe from AI and I will see how many of them I can find information on someone who is already actively working on eliminating them.
Companies don’t want employees, only paying customers. If they can eliminate employees, they will. Hence self-checkouts in grocers, pay at the pump for gas stations, order kiosks at McDonald’s, mobile ordering for virtually every fast food place, the list goes on and on. These are all recent non-AI replacements that have cut into the employment prospects for people.
Pretty sure nah. But time will tell. I will believe it when I see it. AI has been coming for jobs since before terminator. It will replace thousands of jobs just like :
Washer women, lamp lighters, calculators and all the work that farm labourers used to do. Automation comes for us all.
Some jobs shouldn’t exist anyway. God the amount of office workers moving numbers from one tab to another and getting paid a bucket load.
However nursing and elderly care. Psychology counselling mindfulness teachers and jobs that are actually useful for society are probably safe.
Yes ai can do some of all these things but it can’t do them all with empathy. Empathy is key to most of these human focused roles. We need more people in these roles and less working to make more money.
But a lot of jobs did get automated away. And serious consequences did occur from that. Sometimes places rebound from it, but sometimes they did not. And at some point… there will be more people than jobs for them to do, as we continue automating.
In the end, the base foundation for capitalism will be broken, and we will be in an economic crisis of unprecedented scales.
We don’t want to work. We can automate away every job. Then we can be free to actually pursue what we want. Humanity isn’t based on how many shiny trinkets we have.
Yes, but the problem is, we are stuck with the system until we force a societal level change. Capitalism works plenty well enough for the powerful, and they aren’t willing to let go that easily.
“It won’t take people’s jobs! And also people’s jobs are stupid and they deserve to have them taken away!”
What jobs are “useful for society” has no impact on what jobs are actually available to society, only what is deemed “profitable” has any place in this capitalist dystopia. Nice idealism though, I hope it won’t sting too bad having it shattered growing up.
I’m grown up. It will remove jobs. Just said that. Jobs that could be automated regardless. Obviously ai will remove jobs. Just like computers did. But not ones that we actually need. Pretty easy to understand or do you need to grow up to understand that ?
I don’t disagree with most of what you said. I think so far the following jobs are safe from direct AI replacement, because it is much harder to replace manual laborers.
Oil rig worker
Plumber
Construction worker
Landscaper/gardener
Telephone repair tech
Mechanic
Firefighter
Surveyor
Wildlife management officer
Police
What companies won’t realize until too late is that paying customers need jobs to pay for things. If AI causes unemployment to rise to some ungodly high, paying customers will become rare and companies will collapse in droves.
Thanks for actually rising to the challenge, it was actually fascinating to do the research to see how AI is affecting the various industries, and how deeply. I will say that I was able to find direct evidence of replacement in 7/10 of them, 1 was work that is similar and could easily be adapted (telecom line repair), one was an analysis that I think has a lot of good points (plumber), and one was genuinely all about augmenting the capabilities of workers already in place (wildlife conservation/officer).
What companies won’t realize until too late is that paying customers need jobs to pay for things. If AI causes unemployment to rise to some ungodly high, paying customers will become rare and companies will collapse in droves.
I wholeheartedly agree. Functionally, we are going to have to institute a UBI model. It is the only way that society will be able to distribute funds properly when population outpaces jobs due to the exponential growth of populations and the rapidly shrinking landscape of jobs. The corporations are going to need to pay us one way or another.
Damn… nice work on the research! I will read through these as I get time. I genuinely didn’t think there would be much for manual labor stuff. I’m particularly interested in the plumber analysis.
I think augmentation makes a lot of sense for jobs where a human body is needed and it will be interesting to see how/if trade skill requirements change.
I’ll edit this as I read…
Plumbing. The article makes the point that it isn’t all or nothing. That as automation increases productivity, fewer workers are needed. Ok, sure, good point.
Robot plumber? A humanoid robot? Not very likely until enormous breakthroughs are made in machine vision (I can go into more detail…), battery power density, sensor density, etc. The places and situations vary far too greatly.
Rather than an Asimov-style robot, a more feasible yet productivity enhancing solution is automated pipe cutting and other tasks. For example, you go take your phone and measure the pipe as described in the link. Now press a button, walk out to your truck by which time the pipe cutter has already cut off the size you need saving you several minutes. That savings probably means you can do more jobs per day. Cool.
Edit 2
Oil rig worker. Interesting and expected use of AI to improve various aspects of the drilling process. What I had in mind was more like the people that actually do the manual labor.
Autonomous drones, for example, can be used to perform inspections without exposing workers to dangerous situations. In doing so, they can be equipped with sensors that send images and data to operators in real time to enable quick decisions and effective actions for maintenance and repair.
Now that’s pretty cool and will probably reduce demand for those performing inspections (some of whom will have to be at the other end receiving and analyzing data from the robot until such time as AI can do that too.
Autonomous robots, on the other hand, can perform maintenance tasks while making targeted repairs to machinery and equipment.
Again, technologies required to make this happen aren’t there yet. Machine vision (MV) alone is way too far from being general purpose. You can decide a MV system that can, say, detect a coke can and maybe a few other objects under controlled conditions.
But that’s the gotcha.Change the intensity of lighting, change the color temperature or hue of the lighting and the MV probably won’t work. It might also mistake diet coke can or a similar sized cylinder for a Pepsi can. If you want it to recognize any aluminum beverage can that might be tough. Meanwhile any child can easily identify a can in any number of conditions.
Now imagine a diesel engine generator, let’s say. Just getting a robot to change the oil would be nice. But it has to either be limited to a specific model of engine or be able to recognize where the oil drain plug and fill spot is for various engines it might encounter.
What if the engine is a different color? Or dirty instead of clean? Or it’s night, or noon (harsh shadows), overcast (soft shadows), or sunset (everything is yellow orange tinted)? I suppose it could be trained for a specific rig and a specific time of day but that means set up time costs a lot. It might be smarter to build some automated devices on the engine like a valve on the oil pan. And a device to pump new oil in from a vat or standard container or whatever. That would be much easier. Maybe they already do this, idk.
Anyway… progress is being made in MV and we will make far more. That still leaves the question of an autonomous robot of some kind able to remove and reinstall a drain plug. It’s easy for us but you’d be surprised at how hard that would be for a robot.
Had a thought that deserved a separate post. Your selection of MV tasks was rather perverse for the tasks we were discussing. Identifying a pop can is definitely something that humans can do easily because pop cans were made for us to be able to easily identify them. Level the playing field and let’s start looking for internal stress fractures in the superstructure of a 100’ tall concrete bridge. That is something that AI drones are already being designed and deployed for. The drone can easily approach the bridge with a suite of sensors that let it see deep into the superstructure and detect future failure points. Humans would struggle to do that. I have also seen things about maintenance drones that are able to crawl on the bridge using a variety of methods (usually they are designed for specific bridges) that are able to fill cracks with sealant and ablate rust using lasers, then paint the freshly cleaned metal. The benefit of replacing a workforce with AI-driven robotics is that you can purpose-build and purpose-train the tool to do exactly what you need it to do. A robot that scurries into a crawl space to run a pipe for a plumber doesn’t need to know how to do anything but recognize where it goes, what not to touch, and how much force to use when installing it. It doesn’t need to identify a pop can, it doesn’t need to draw a Rembrandt. All it needs to do is pull a pipe and weld it in place (and yes, I am oversimplifying a bit, I know that).
The other thing that kinda gets me is the whole “cramped spaces” safety net that I kept seeing for why this job or that was going to be safe. Designing a small, agile robot is not really a challenge. Add onto it that in many situations you could use a tethered drone to do the actual work that is much smaller and the AI brain can be sitting safely outside the situation. You could even plug it into power, so battery tech doesn’t need to increase. shrug I guess I just see quite a bit of very fast advances in the tech that have a worrying trajectory to me.
All great points. I guess I need to think of this topic more from the “what is possible” mindset rather than the “this is too hard” mindset to get a fair assessment of what is coming. All while still framing it in the sense of improving worker efficiency and automating human tasks piecemeal over time.
Agree that other parts of the EM spectrum could enhance the ability of MV to recognize things. Appreciate the insights – maybe I will be able to use this when I get back to tinkering with MV as a hobbyist.
Of course identifying one object is one level. For a general purpose replacement for humans ability, since that’s what the thread is focused (ahem) on, it has to identify tens of thousands of objects.
I need to rethink my opinion a bit. Not only how far general object recognition is but also how one can “cheat” to enable robotic automation.
Tasks that are more limited in scope and variability would be a lot less demanding. For a silly example, let’s say we want to automate replacing fuses in cars. We limit it to cars with fuse boxes in the engine bay and we can mark the fuse box with a visual tag the robot can detect. The layout of the fuses per vehicle model could be stored. The code on the fuse box identifies the model. The robot then used actuators to remove the cover and orients itself to the box using more markers and the rest is basically pick and place technology. That’s a smaller and easier problem to solve than “fix anything possibly wrong with a car”. A similar deal could be done for oil changes.
For general purpose MV object detection, I would have to go check but my guess is that what is possible with state of the art MV is identifying a dozen or maybe even hundreds of objects so I suppose one could do quite a bit with that to automate some jobs. MV is not to my knowledge at a level of general purpose replacement for humans. Yet. Maybe it won’t take that much longer.
In ~15 years in the hobbyist space we’ve gone from recognizing anything of a specified color under some lighting conditions to identifying several specific objects. And without a ton of processing power either. It’s pretty damn impressive progress, really. We have security cameras that can identify animals, people, and delivery boxes. I am probably selling short what MV will be able to do in 15 more years.
This is part of the problem. They don’t, and won’t, fully understand the technology or its limitations or long-term impacts. They will understand that the salesman pushing the AI product told them it could eliminate 5-10% of their workforce. Whether or not the product can actually do that effectively won’t matter, they’ll still buy it, implement it, and fire a bunch of people.
I think once sap and jira start implementing a lot more AI and make it simpler to use it could cut down a lot of corporate jobs, not the hands on stuff but a lot of the simpler jobs like purchasing and inventory staff could be shrunken down to a fewer people and fewer cubicles. At least that’s what we talked about at our company how everyone is adjusting to the new world especially advertising now that everything will be served to you by a bot instead of a search
Do you believe this LLM tech has the ability to make judgement calls, say? Or understand meaning?
What has been your experience with the accuracy / correctness of the answers it has provided? Does it match claims that mistakes or “hallucinations” occur often?
You’re wandering into one of the great questions of our age: what is intelligence? I don’t have a great answer. All I know is that gpt-4 can REASON, and does so better than the average human.
It’s gpt-4 self-aware? Yes. To an extent. It knows what it is, and can use that information in its reasoning. It knows it’s an LLM, but not which model.
Can it make judgement calls? Yes. Better than the average human.
Understand meaning? Absolutely. To a jaw-dropping extent.
Accuracy and correctness… Depends on the type of question.
What you need to understand is that gpt-4 isn’t a whole brain. Think of it as if we have managed to reproduce the language center of the brain. I believe this is mechanism for higher reasoning in the human brain.
But just as in humans with right-brain injuries, the language center is disconnected from reality at times.
So, when you think about gpt-4 as the most important, difficult to solve part of the brain, you start to understand that with some minimal supporting infrastructure, you now have something very similar to a complete brain.
You can use vector databases to give it long-term memory, and any kind of data retrieval used to augment it’s prompts improved accuracy and reduces hallucinations almost entirely.
With my very mediocre programming skills, I managed to build a system that is curious, has a long-term memory, and do a wide variety of tasks, enough to easily replace an entire customer service, tech support team, sales team, and marketing team.
That’s just ME, and working with the gpt-4 that’s available to the public with a bunch of guardrails on it. Today.
Imagine a less-restricted system, with infrastructure built by an experienced enterprise coding team, and with just one more generation of LLM improvement? That could wipe out half the white collar workforce.
If LLM improvement was only geometric, and not even exponential (as it clearly is), in 10 years these things will be smarter AND MORE CREATIVE than all humans.
The truth is that we’re going to be there in 5 years.
Indeed, intelligence is …a difficult thing to define. It’s also a fascinating area to ponder. The reason I asked was to get an idea of where your head is at with the claims you made.
Now, I admit I haven’t done a lot with gpt-4 but your comments make me think it is worth the time to do so.
So you indicate gpt-4 can reason. My understanding is gpt-4 is an LLM, basically a large scale Markov chain, trained to respond with appropriate output based on input (questions).
On the one hand, my initial reaction is: no, it doesn’t reason it just mimics or simulates human reasoning that came before it in text form.
On the other hand, if a program could perfectly simulate whatever processes are involved in reasoning by a human to the point that they’re indistinguishable, is it not, in effect, reasoning? (I suppose this amounts to a sort of Turing Test but for reasoning exercises).
I don’t know how gpt4 LLMs work yet. I imagine, being a Markov Model (specifically a Markov Chain), if the model is trained on human language then the underlying semantics are sort of implicitly captured in the statistical model. Like, simplistically, if many sentences reflect human knowledge that cars are vehicles and not animals then it’s statistically unlikely for anyone to write about attributes and actions of animals when talking about cars. I assume the LLM is of such a scale that it permits this apparently emergent behavior.
I am skeptical about judgement calls. I would think some sensory input would be required. I guess we have to outline various types of judgement calls to really dig into this.
I am willing to accept that gpt-4 simulates the portions of the brain that deal with semantics and syntax both the receiving and transmitting abilities. And, maybe to some degree, knowledge and understanding.
I think “very similar to a complete brain” is an overstatement as the brain also does some amazing things with vision, hearing, proprioception, touch, among other things. Human brains can analyze situations and take initiative, analyze things and understand how they work and apply that to their repair, improvement, duplication, etc. We can understand and solve problems, and so on. In other words I don’t think you’re giving the brain anywhere near enough credit. We aren’t just Q&A machines.
We also have to be careful of the human tendency to anthropomorphize.
I’m curious to look into vector databases and their applications here. Addition of what amounts to memory, or like extended context, sounds extremely interesting.
Interesting to ponder what the world would be like with AGI taking over the jobs of most knowledge workers, artists, and so on. (I wonder if someone could create a CEO replacement…)
What does it mean for a capitalist society with masses of people permanently unemployed? How does the economy work when nobody can afford to buy anything because they’re unemployed? Does this create widespread poverty and collapse or a post-scarcity economy in some sectors?
Until robots mechanically evolve to Asimov’s vision, at least, manual labor is safe. Truly being able to replace a human body with a robot is still a ways off due to lack of progress on several fronts.
You sound like one of those peasants standing on street corners saying, “horses replaced with fuming metal boxes in 10 years? Hah, yeah, sure buddy, right after we put a man on the moon! Getoutta here, you loon!”
There is a video from CGP Grey titled Humans Need Not Apply that is extremely relevant. It was posted 9 years ago. It’s a great video, I highly recommend everyone check it out.
Thanks for sharing. If you see that list of type of jobs at the end, it’s easy to see which jobs could get replaced within a reasonably short amount of time.
Greed will always find a way to profit from whatever development arises. If they have 1 mountain of gold, they want 2 mountains of gold.
Yup. This is why it is vital that we all get behind Universal Basic Income.
The jobs will leave and they won’t come back. UBI is inevitable, but if we don’t get there soon enough there will be years of suffering and poverty for hundreds of millions.
I’m a senior Linux sysadmin who’s been following the evolution of AI over this past year just like you, and just like you I’ve been spending my days and nights tinkering with it non stop, and I have come to more or less the same conclusion as you have.
The downvotes are from people who haven’t used the AI, and who are still in the Internet 1.0 mindset. How people still don’t get just how revolutionary this technology is, is beyond me. But yeah, in a few years that’ll be evident enough, time will show.
This is just the smallest tip of the iceberg.
I’ve been working with gpt-4 since the week it came out, and I guarantee you that even if it never became any more advanced, it could already put at least 30% of the white collar workforce out of business.
The only reason it hasn’t is because companies have barely started to comprehend what it can do.
Within 5 years the entire world will have been revolutionized by this technology. Jobs will evaporate faster than anyone is talking about.
If you’re very smart, and you begin to use gpt-4 to write the tools that will replace you, then you MIGHT have 10 good years left in this economy before humans are all but obsolete.
If you’re not staying up nights, scared shitless by what’s coming, it’s because you don’t really understand what gpt-4 can do.
You sound like one of those idiots preaching the apocalypse from a street corner. Humans obsolete in 10 years? Yeah sure buddy, right after all those profits trickle down. This is just another tool, an interesting one to be sure, but still just a tool. If you’re staying up nights worrying about this, you don’t really understand the technology, or maybe you’re just worried someone is going to realize you don’t do shit.
I work with AI stuff, just getting into LLM, but I have been doing SD work since the public release last year. In just over 1 year the SD capability has gone from being able to draw a passable image of a cat at 512x512 pixels that required a reasonably powerful graphics card to complete to being able to create 4k images on the same cards that are nearly indistinguishable from actual photos/paintings. It is the single fastest adaptation and development of a technology I have seen in my 30 years in tech. I have actually been tracking the job market and the impacts that this will have and he is not all that far off in his estimate. The current push in AI development is nearly a ubiquitous existential threat to employment as we view it in the society of the United States. Everyone is on the chopping block and you’d best believe that the C-level executives want to eliminate as many positions as possible. Labor is viewed as an atrocious expense and the first place that cuts should be made. I challenge you to actually come up with a list of 10 jobs that employ more than 100,000 people in the country that you think would be safe from AI and I will see how many of them I can find information on someone who is already actively working on eliminating them.
Companies don’t want employees, only paying customers. If they can eliminate employees, they will. Hence self-checkouts in grocers, pay at the pump for gas stations, order kiosks at McDonald’s, mobile ordering for virtually every fast food place, the list goes on and on. These are all recent non-AI replacements that have cut into the employment prospects for people.
Pretty sure nah. But time will tell. I will believe it when I see it. AI has been coming for jobs since before terminator. It will replace thousands of jobs just like :
Washer women, lamp lighters, calculators and all the work that farm labourers used to do. Automation comes for us all.
Some jobs shouldn’t exist anyway. God the amount of office workers moving numbers from one tab to another and getting paid a bucket load.
However nursing and elderly care. Psychology counselling mindfulness teachers and jobs that are actually useful for society are probably safe. Yes ai can do some of all these things but it can’t do them all with empathy. Empathy is key to most of these human focused roles. We need more people in these roles and less working to make more money.
But a lot of jobs did get automated away. And serious consequences did occur from that. Sometimes places rebound from it, but sometimes they did not. And at some point… there will be more people than jobs for them to do, as we continue automating.
In the end, the base foundation for capitalism will be broken, and we will be in an economic crisis of unprecedented scales.
Capitalism doesn’t work. Pretty sure everyone knows that.
We don’t want to work. We can automate away every job. Then we can be free to actually pursue what we want. Humanity isn’t based on how many shiny trinkets we have.
Yes, but the problem is, we are stuck with the system until we force a societal level change. Capitalism works plenty well enough for the powerful, and they aren’t willing to let go that easily.
I don’t disagree. Bring on the revolution
“It won’t take people’s jobs! And also people’s jobs are stupid and they deserve to have them taken away!”
What jobs are “useful for society” has no impact on what jobs are actually available to society, only what is deemed “profitable” has any place in this capitalist dystopia. Nice idealism though, I hope it won’t sting too bad having it shattered growing up.
I’m grown up. It will remove jobs. Just said that. Jobs that could be automated regardless. Obviously ai will remove jobs. Just like computers did. But not ones that we actually need. Pretty easy to understand or do you need to grow up to understand that ?
I don’t disagree with most of what you said. I think so far the following jobs are safe from direct AI replacement, because it is much harder to replace manual laborers.
What companies won’t realize until too late is that paying customers need jobs to pay for things. If AI causes unemployment to rise to some ungodly high, paying customers will become rare and companies will collapse in droves.
Thanks for actually rising to the challenge, it was actually fascinating to do the research to see how AI is affecting the various industries, and how deeply. I will say that I was able to find direct evidence of replacement in 7/10 of them, 1 was work that is similar and could easily be adapted (telecom line repair), one was an analysis that I think has a lot of good points (plumber), and one was genuinely all about augmenting the capabilities of workers already in place (wildlife conservation/officer).
I wholeheartedly agree. Functionally, we are going to have to institute a UBI model. It is the only way that society will be able to distribute funds properly when population outpaces jobs due to the exponential growth of populations and the rapidly shrinking landscape of jobs. The corporations are going to need to pay us one way or another.
Damn… nice work on the research! I will read through these as I get time. I genuinely didn’t think there would be much for manual labor stuff. I’m particularly interested in the plumber analysis.
I think augmentation makes a lot of sense for jobs where a human body is needed and it will be interesting to see how/if trade skill requirements change.
I’ll edit this as I read…
Plumbing. The article makes the point that it isn’t all or nothing. That as automation increases productivity, fewer workers are needed. Ok, sure, good point.
Robot plumber? A humanoid robot? Not very likely until enormous breakthroughs are made in machine vision (I can go into more detail…), battery power density, sensor density, etc. The places and situations vary far too greatly.
Rather than an Asimov-style robot, a more feasible yet productivity enhancing solution is automated pipe cutting and other tasks. For example, you go take your phone and measure the pipe as described in the link. Now press a button, walk out to your truck by which time the pipe cutter has already cut off the size you need saving you several minutes. That savings probably means you can do more jobs per day. Cool.
Edit 2
Oil rig worker. Interesting and expected use of AI to improve various aspects of the drilling process. What I had in mind was more like the people that actually do the manual labor.
Now that’s pretty cool and will probably reduce demand for those performing inspections (some of whom will have to be at the other end receiving and analyzing data from the robot until such time as AI can do that too.
Again, technologies required to make this happen aren’t there yet. Machine vision (MV) alone is way too far from being general purpose. You can decide a MV system that can, say, detect a coke can and maybe a few other objects under controlled conditions.
But that’s the gotcha.Change the intensity of lighting, change the color temperature or hue of the lighting and the MV probably won’t work. It might also mistake diet coke can or a similar sized cylinder for a Pepsi can. If you want it to recognize any aluminum beverage can that might be tough. Meanwhile any child can easily identify a can in any number of conditions.
Now imagine a diesel engine generator, let’s say. Just getting a robot to change the oil would be nice. But it has to either be limited to a specific model of engine or be able to recognize where the oil drain plug and fill spot is for various engines it might encounter.
What if the engine is a different color? Or dirty instead of clean? Or it’s night, or noon (harsh shadows), overcast (soft shadows), or sunset (everything is yellow orange tinted)? I suppose it could be trained for a specific rig and a specific time of day but that means set up time costs a lot. It might be smarter to build some automated devices on the engine like a valve on the oil pan. And a device to pump new oil in from a vat or standard container or whatever. That would be much easier. Maybe they already do this, idk.
Anyway… progress is being made in MV and we will make far more. That still leaves the question of an autonomous robot of some kind able to remove and reinstall a drain plug. It’s easy for us but you’d be surprised at how hard that would be for a robot.
Had a thought that deserved a separate post. Your selection of MV tasks was rather perverse for the tasks we were discussing. Identifying a pop can is definitely something that humans can do easily because pop cans were made for us to be able to easily identify them. Level the playing field and let’s start looking for internal stress fractures in the superstructure of a 100’ tall concrete bridge. That is something that AI drones are already being designed and deployed for. The drone can easily approach the bridge with a suite of sensors that let it see deep into the superstructure and detect future failure points. Humans would struggle to do that. I have also seen things about maintenance drones that are able to crawl on the bridge using a variety of methods (usually they are designed for specific bridges) that are able to fill cracks with sealant and ablate rust using lasers, then paint the freshly cleaned metal. The benefit of replacing a workforce with AI-driven robotics is that you can purpose-build and purpose-train the tool to do exactly what you need it to do. A robot that scurries into a crawl space to run a pipe for a plumber doesn’t need to know how to do anything but recognize where it goes, what not to touch, and how much force to use when installing it. It doesn’t need to identify a pop can, it doesn’t need to draw a Rembrandt. All it needs to do is pull a pipe and weld it in place (and yes, I am oversimplifying a bit, I know that).
The other thing that kinda gets me is the whole “cramped spaces” safety net that I kept seeing for why this job or that was going to be safe. Designing a small, agile robot is not really a challenge. Add onto it that in many situations you could use a tethered drone to do the actual work that is much smaller and the AI brain can be sitting safely outside the situation. You could even plug it into power, so battery tech doesn’t need to increase. shrug I guess I just see quite a bit of very fast advances in the tech that have a worrying trajectory to me.
All great points. I guess I need to think of this topic more from the “what is possible” mindset rather than the “this is too hard” mindset to get a fair assessment of what is coming. All while still framing it in the sense of improving worker efficiency and automating human tasks piecemeal over time.
Your points on MV are not unfounded, but they are also extremely homeocentric. All of your examples rely on the visible light spectrum as well as standard “vision” as we know it. Realistically any sensor can be used to generate an image if you know what you are doing with it. Radio telescopes are a great example of this. There is also a lot of research going on in giving AI’s MV senses access to other sections of the EM spectrum ( https://www.edge-ai-vision.com/2017/10/beyond-visible-light-applications-in-computer-vision/ and https://www.technologyreview.com/2019/10/09/132696/machine-vision-has-learned-to-use-radio-waves-to-see-through-walls-and-in-darkness/ ) as well as echolocation ( https://www.imveurope.com/news/echolocation-neural-net-gives-phones-3d-vision-sound ). There are many other types of “vision” that can be used that can definitely distinguish a popcan.
Agree that other parts of the EM spectrum could enhance the ability of MV to recognize things. Appreciate the insights – maybe I will be able to use this when I get back to tinkering with MV as a hobbyist.
Of course identifying one object is one level. For a general purpose replacement for humans ability, since that’s what the thread is focused (ahem) on, it has to identify tens of thousands of objects.
I need to rethink my opinion a bit. Not only how far general object recognition is but also how one can “cheat” to enable robotic automation.
Tasks that are more limited in scope and variability would be a lot less demanding. For a silly example, let’s say we want to automate replacing fuses in cars. We limit it to cars with fuse boxes in the engine bay and we can mark the fuse box with a visual tag the robot can detect. The layout of the fuses per vehicle model could be stored. The code on the fuse box identifies the model. The robot then used actuators to remove the cover and orients itself to the box using more markers and the rest is basically pick and place technology. That’s a smaller and easier problem to solve than “fix anything possibly wrong with a car”. A similar deal could be done for oil changes.
For general purpose MV object detection, I would have to go check but my guess is that what is possible with state of the art MV is identifying a dozen or maybe even hundreds of objects so I suppose one could do quite a bit with that to automate some jobs. MV is not to my knowledge at a level of general purpose replacement for humans. Yet. Maybe it won’t take that much longer.
In ~15 years in the hobbyist space we’ve gone from recognizing anything of a specified color under some lighting conditions to identifying several specific objects. And without a ton of processing power either. It’s pretty damn impressive progress, really. We have security cameras that can identify animals, people, and delivery boxes. I am probably selling short what MV will be able to do in 15 more years.
And you think managers, the people deciding who gets replaced by AI, understand the technology?
This is part of the problem. They don’t, and won’t, fully understand the technology or its limitations or long-term impacts. They will understand that the salesman pushing the AI product told them it could eliminate 5-10% of their workforce. Whether or not the product can actually do that effectively won’t matter, they’ll still buy it, implement it, and fire a bunch of people.
I think once sap and jira start implementing a lot more AI and make it simpler to use it could cut down a lot of corporate jobs, not the hands on stuff but a lot of the simpler jobs like purchasing and inventory staff could be shrunken down to a fewer people and fewer cubicles. At least that’s what we talked about at our company how everyone is adjusting to the new world especially advertising now that everything will be served to you by a bot instead of a search
As I said, if you’re not scared shitless, you really don’t understand what gpt-4 can do.
It’s not “just another tool.”
It’s an intelligence.
This technology is more world-changing than computers, the Internet, or mobile technology. And it’s evolving faster than any of those things did.
You’ll see. Unfortunately.
How do you define “intelligence” in this context?
Do you think gpt4 is self aware?
Do you believe this LLM tech has the ability to make judgement calls, say? Or understand meaning?
What has been your experience with the accuracy / correctness of the answers it has provided? Does it match claims that mistakes or “hallucinations” occur often?
You’re wandering into one of the great questions of our age: what is intelligence? I don’t have a great answer. All I know is that gpt-4 can REASON, and does so better than the average human.
It’s gpt-4 self-aware? Yes. To an extent. It knows what it is, and can use that information in its reasoning. It knows it’s an LLM, but not which model.
Can it make judgement calls? Yes. Better than the average human.
Understand meaning? Absolutely. To a jaw-dropping extent.
Accuracy and correctness… Depends on the type of question.
What you need to understand is that gpt-4 isn’t a whole brain. Think of it as if we have managed to reproduce the language center of the brain. I believe this is mechanism for higher reasoning in the human brain.
But just as in humans with right-brain injuries, the language center is disconnected from reality at times.
So, when you think about gpt-4 as the most important, difficult to solve part of the brain, you start to understand that with some minimal supporting infrastructure, you now have something very similar to a complete brain.
You can use vector databases to give it long-term memory, and any kind of data retrieval used to augment it’s prompts improved accuracy and reduces hallucinations almost entirely.
With my very mediocre programming skills, I managed to build a system that is curious, has a long-term memory, and do a wide variety of tasks, enough to easily replace an entire customer service, tech support team, sales team, and marketing team.
That’s just ME, and working with the gpt-4 that’s available to the public with a bunch of guardrails on it. Today.
Imagine a less-restricted system, with infrastructure built by an experienced enterprise coding team, and with just one more generation of LLM improvement? That could wipe out half the white collar workforce.
If LLM improvement was only geometric, and not even exponential (as it clearly is), in 10 years these things will be smarter AND MORE CREATIVE than all humans.
The truth is that we’re going to be there in 5 years.
Appreciate the detailed response!
Indeed, intelligence is …a difficult thing to define. It’s also a fascinating area to ponder. The reason I asked was to get an idea of where your head is at with the claims you made.
Now, I admit I haven’t done a lot with gpt-4 but your comments make me think it is worth the time to do so.
So you indicate gpt-4 can reason. My understanding is gpt-4 is an LLM, basically a large scale Markov chain, trained to respond with appropriate output based on input (questions).
On the one hand, my initial reaction is: no, it doesn’t reason it just mimics or simulates human reasoning that came before it in text form.
On the other hand, if a program could perfectly simulate whatever processes are involved in reasoning by a human to the point that they’re indistinguishable, is it not, in effect, reasoning? (I suppose this amounts to a sort of Turing Test but for reasoning exercises).
I don’t know how gpt4 LLMs work yet. I imagine, being a Markov Model (specifically a Markov Chain), if the model is trained on human language then the underlying semantics are sort of implicitly captured in the statistical model. Like, simplistically, if many sentences reflect human knowledge that cars are vehicles and not animals then it’s statistically unlikely for anyone to write about attributes and actions of animals when talking about cars. I assume the LLM is of such a scale that it permits this apparently emergent behavior.
I am skeptical about judgement calls. I would think some sensory input would be required. I guess we have to outline various types of judgement calls to really dig into this.
I am willing to accept that gpt-4 simulates the portions of the brain that deal with semantics and syntax both the receiving and transmitting abilities. And, maybe to some degree, knowledge and understanding.
I think “very similar to a complete brain” is an overstatement as the brain also does some amazing things with vision, hearing, proprioception, touch, among other things. Human brains can analyze situations and take initiative, analyze things and understand how they work and apply that to their repair, improvement, duplication, etc. We can understand and solve problems, and so on. In other words I don’t think you’re giving the brain anywhere near enough credit. We aren’t just Q&A machines.
We also have to be careful of the human tendency to anthropomorphize.
I’m curious to look into vector databases and their applications here. Addition of what amounts to memory, or like extended context, sounds extremely interesting.
Interesting to ponder what the world would be like with AGI taking over the jobs of most knowledge workers, artists, and so on. (I wonder if someone could create a CEO replacement…)
What does it mean for a capitalist society with masses of people permanently unemployed? How does the economy work when nobody can afford to buy anything because they’re unemployed? Does this create widespread poverty and collapse or a post-scarcity economy in some sectors?
Until robots mechanically evolve to Asimov’s vision, at least, manual labor is safe. Truly being able to replace a human body with a robot is still a ways off due to lack of progress on several fronts.
You sound like one of those peasants standing on street corners saying, “horses replaced with fuming metal boxes in 10 years? Hah, yeah, sure buddy, right after we put a man on the moon! Getoutta here, you loon!”
There is a video from CGP Grey titled Humans Need Not Apply that is extremely relevant. It was posted 9 years ago. It’s a great video, I highly recommend everyone check it out.
Here is an alternative Piped link(s):
Humans Need Not Apply
Piped is a privacy-respecting open-source alternative frontend to YouTube.
I’m open-source; check me out at GitHub.
Thanks for sharing. If you see that list of type of jobs at the end, it’s easy to see which jobs could get replaced within a reasonably short amount of time. Greed will always find a way to profit from whatever development arises. If they have 1 mountain of gold, they want 2 mountains of gold.
Yup. This is why it is vital that we all get behind Universal Basic Income.
The jobs will leave and they won’t come back. UBI is inevitable, but if we don’t get there soon enough there will be years of suffering and poverty for hundreds of millions.
I’m a senior Linux sysadmin who’s been following the evolution of AI over this past year just like you, and just like you I’ve been spending my days and nights tinkering with it non stop, and I have come to more or less the same conclusion as you have.
The downvotes are from people who haven’t used the AI, and who are still in the Internet 1.0 mindset. How people still don’t get just how revolutionary this technology is, is beyond me. But yeah, in a few years that’ll be evident enough, time will show.
I feel sorry for these folks. They have no idea what’s about to happen.