The AI boom has turned the standard profit margin model on its head, according to Apollo Chief Economist Torsten Slok—and it’s making the industry’s growth unsustainable.
When they invented the microscope, hey could instantly demonstrate its utility in it’s ability to observe previously invisibly tiny things.
I assure you that 400 some odd years ago, most people would not have grasped the utility in the microscope. Just like many people today don’t understand what the transformer has done for machine learning.
But some could, and they could articulate their reasoning. “Machine learning” is not even necessarily a path to anything. Its hard to argue that against your example, which is literally unlocking secrets of the universe.
Again, you’re using 400+ years of hindsight to inform your conclusion. “Looking at small things” doesn’t immediately imply “unlocking the secrets of the universe”, it’s only after hundreds of years of building on that invention that you can easily point to that conclusion.
And admittedly the analogy was not great, I’m just trying to get across that you’re looking at the output of the invention and asking “is that it”? But we’ve only had this technology for a decade or so, and it’s only really been public facing for, what, 4 years?
These various technologies we refer to as “AI” are not yet at the point where they can replace any human at any task, but they can replace some humans at some tasks, and as far as the evidence shows, we can continue to improve it with more training and more compute. They are writing prose well enough to pass as humans, solving math problems we haven’t been able to solve for decades, writing medium complexity code in minutes instead of days-- the technology isn’t solving all the world’s problems and ushering in a post-scarcity utopia, but no one is claiming it is, today. Will it? I doubt it, but that’s more to do with capitalism corrupting everything it touches than anything else.
If you think “AI” is not useful, you’re probably considering a very narrowly defined type of “AI” in a very narrowly defined scenario where AI does not perform well.
Edit: I want to take a second to say that I’ve enjoyed this conversation with you. I can’t say that about many conversations on this topic. This isn’t me ending the conversation.
I’m not even drawing a conclusion anymore. I’m asking for anyone anywhere to point to the previously impossible thing that AI lets me do. Even if that thing is of questionable utility, what is it? I just want an answer to that question.
Then, we can start talking about if that’s worth the cost. Most previous inventions cost someone’s life savings, or corporate focus on technology. This one’s costing us the GDP of South Korea, so it better be good. Like “the microscope” or “the wheel” good.
Everyone’s claiming that! Every invention allows someone to do something that was previously impossible. Humans couldn’t travel faster than a horse can run until the car was invented. Humans couldn’t see microbes before the microscope was invented. People couldn’t drag more than 10x their body weight until the wheel was invented. People couldn’t catch a marlin until the net was invented.
I am trying to resist the notion that youre being willfully obtuse if you aren’t equating invention with increasing what is possible.
I didn’t realize you meant that. In my mind humans could drag stuff and catch fish but the inventions made it easier or more efficient.
In that view, pretty much anything on a computer can be done easier or more efficient with AI. Same as a net or a wheel. Just Google anything it is you might want to do on a computer and add “using AI” and you’re likely to find a tool to do it. If you have the computer for it, you can download an open weight model to do locally (though probably not as well as using the crazy compute associated with using an API model.
Also, as agentic AI improves, we’ll see the scarcity of attention decrease and maybe vanish entirely. It’s not always obvious, but much of how our society is structured assuming that people have limited agency.
I do see what you mean. You’re saying that AI allows you to do something much faster or perform more work in the same time by essentially doing parts of it for you. While that is true, it is only so for certain values of “more”. For example, I could get an AI to write most or all of a configuration for a server deployment for me in about 2 minutes, instead of maybe an hour. However, if I do not spend about an hour reviewing its work, I will get a failure rate dramatically higher than if I had just done it myself. Similarly, if I decide to start pencil-whipping my work to do many times more work in the same time, I could do that and get much more deliverable work out the door. Whether or not you would call that more productive depends on what factors you’re accounting for. So, for applications I am familiar with, the primary value proposition of LLMs is that I can maximize quantity over quality even more than would be possible without it, by allowing me to not even do much of it. The overall quality of the output will likely improve incrementally over time as it has so far, but given its current design, its propensity for errors and hallucinations can only be mitigated, never eliminated or made trivial.
Because of this limitation inherent in its design, I don’t believe it can ever have real applications for tasks where quality is important. That said, there are many tasks where it is not. For example, fuzzy correlations run over vast data sets can be riddled with false correlations and still be useful, like Flock does. It is also effective at filling “gaps” in human knowledge, such as a mathematical truth that we have yet to devise a mathematical proof for, or structural designs that a human could come up with, but never did because of how unintuitive the solution is to humans. Basically, any problem that can be solved by slinging tons of shit at the wall to see what sticks is fertile ground for AIs to help us.
I, personally, enjoy creating something much more than I enjoy proofreading someone else’s creation for the same amount of time. I, and I’m sure many, many others, don’t want their 40 hour work week of designing things and solving problems to be replaced with 40 hours of proofreading and debugging to achieve a comparable amount of productivity. Once again, the proof is in the pudding. Tech workers (and likely all workers) like tools that make their lives easier. If these tools were doing that, our bosses would not need to force us to use them; we would want to use them so we can go home early. I just simply don’t agree that they let most workers get more done in less time, and I cite the fact that nobody is, when you actually measure it rather than going on feeling, getting more done in less time. If this were happening anywhere in any numbers, these AI marketers would be shouting it from the rooftops. As mentioned above, there are exceptions to that, but they are just that, exceptions.
I assure you that 400 some odd years ago, most people would not have grasped the utility in the microscope. Just like many people today don’t understand what the transformer has done for machine learning.
But some could, and they could articulate their reasoning. “Machine learning” is not even necessarily a path to anything. Its hard to argue that against your example, which is literally unlocking secrets of the universe.
Again, you’re using 400+ years of hindsight to inform your conclusion. “Looking at small things” doesn’t immediately imply “unlocking the secrets of the universe”, it’s only after hundreds of years of building on that invention that you can easily point to that conclusion.
And admittedly the analogy was not great, I’m just trying to get across that you’re looking at the output of the invention and asking “is that it”? But we’ve only had this technology for a decade or so, and it’s only really been public facing for, what, 4 years?
These various technologies we refer to as “AI” are not yet at the point where they can replace any human at any task, but they can replace some humans at some tasks, and as far as the evidence shows, we can continue to improve it with more training and more compute. They are writing prose well enough to pass as humans, solving math problems we haven’t been able to solve for decades, writing medium complexity code in minutes instead of days-- the technology isn’t solving all the world’s problems and ushering in a post-scarcity utopia, but no one is claiming it is, today. Will it? I doubt it, but that’s more to do with capitalism corrupting everything it touches than anything else.
If you think “AI” is not useful, you’re probably considering a very narrowly defined type of “AI” in a very narrowly defined scenario where AI does not perform well.
Edit: I want to take a second to say that I’ve enjoyed this conversation with you. I can’t say that about many conversations on this topic. This isn’t me ending the conversation.
I’m not even drawing a conclusion anymore. I’m asking for anyone anywhere to point to the previously impossible thing that AI lets me do. Even if that thing is of questionable utility, what is it? I just want an answer to that question.
Then, we can start talking about if that’s worth the cost. Most previous inventions cost someone’s life savings, or corporate focus on technology. This one’s costing us the GDP of South Korea, so it better be good. Like “the microscope” or “the wheel” good.
When did anyone ever claim that? It’s not going to break the laws of physics.
If you do it on a computer, current AI (an umbrella term) can probably make it easier for you to do, if not just do it for you.
Everyone’s claiming that! Every invention allows someone to do something that was previously impossible. Humans couldn’t travel faster than a horse can run until the car was invented. Humans couldn’t see microbes before the microscope was invented. People couldn’t drag more than 10x their body weight until the wheel was invented. People couldn’t catch a marlin until the net was invented.
I am trying to resist the notion that youre being willfully obtuse if you aren’t equating invention with increasing what is possible.
I didn’t realize you meant that. In my mind humans could drag stuff and catch fish but the inventions made it easier or more efficient.
In that view, pretty much anything on a computer can be done easier or more efficient with AI. Same as a net or a wheel. Just Google anything it is you might want to do on a computer and add “using AI” and you’re likely to find a tool to do it. If you have the computer for it, you can download an open weight model to do locally (though probably not as well as using the crazy compute associated with using an API model.
Also, as agentic AI improves, we’ll see the scarcity of attention decrease and maybe vanish entirely. It’s not always obvious, but much of how our society is structured assuming that people have limited agency.
Do you mind sharing what your profession is?
I do see what you mean. You’re saying that AI allows you to do something much faster or perform more work in the same time by essentially doing parts of it for you. While that is true, it is only so for certain values of “more”. For example, I could get an AI to write most or all of a configuration for a server deployment for me in about 2 minutes, instead of maybe an hour. However, if I do not spend about an hour reviewing its work, I will get a failure rate dramatically higher than if I had just done it myself. Similarly, if I decide to start pencil-whipping my work to do many times more work in the same time, I could do that and get much more deliverable work out the door. Whether or not you would call that more productive depends on what factors you’re accounting for. So, for applications I am familiar with, the primary value proposition of LLMs is that I can maximize quantity over quality even more than would be possible without it, by allowing me to not even do much of it. The overall quality of the output will likely improve incrementally over time as it has so far, but given its current design, its propensity for errors and hallucinations can only be mitigated, never eliminated or made trivial.
Because of this limitation inherent in its design, I don’t believe it can ever have real applications for tasks where quality is important. That said, there are many tasks where it is not. For example, fuzzy correlations run over vast data sets can be riddled with false correlations and still be useful, like Flock does. It is also effective at filling “gaps” in human knowledge, such as a mathematical truth that we have yet to devise a mathematical proof for, or structural designs that a human could come up with, but never did because of how unintuitive the solution is to humans. Basically, any problem that can be solved by slinging tons of shit at the wall to see what sticks is fertile ground for AIs to help us.
I, personally, enjoy creating something much more than I enjoy proofreading someone else’s creation for the same amount of time. I, and I’m sure many, many others, don’t want their 40 hour work week of designing things and solving problems to be replaced with 40 hours of proofreading and debugging to achieve a comparable amount of productivity. Once again, the proof is in the pudding. Tech workers (and likely all workers) like tools that make their lives easier. If these tools were doing that, our bosses would not need to force us to use them; we would want to use them so we can go home early. I just simply don’t agree that they let most workers get more done in less time, and I cite the fact that nobody is, when you actually measure it rather than going on feeling, getting more done in less time. If this were happening anywhere in any numbers, these AI marketers would be shouting it from the rooftops. As mentioned above, there are exceptions to that, but they are just that, exceptions.
It should definitely not take the same amount of time to check someone’s work as it does to do the work.