Thanks Alberto - right on and very timely. PLEASE Amplify and clarify what you mean in #15.
I am stumped by it. What aother "timelines" and "AI Alignment" are your referring to? As a student of the impact of technologoy and innovatiopn on work, professions, and careers across industries over 4 decades, you nail it. Esp #12!
12. We should transition from “AI is eating the workforce” to “AI has created a new workforce” with safety nets and governmental help available to those who need it.
Thank you Matt! With timelines I refer to "what could have happened" scenarios (counterfactuals). With AI alignment I recommend starting on Wikipedia haha there's really a lot to read on that
In #15 - when you wrote, "There's a limited number of timelines where your job is the biggest of your worries:" What are you referring to? This is what I am confused about. The typical working life of adults 20's-70's is still a lifetime.
Sounds ominous "...and you may be dead if AI alignment is not solved." Are you refererring to significant events when relying on AI was inapproriate or misguided e.g. the failure of the weapon detection system @ Antioch HS and other failures? or somthing else. Thanks again.
The coolest thing about this post was that it really sounds like it wasn't generated by an AI - it's far more insightful than the average content on the internet that LLMs are trained on :)
Thank you Oliver. It absolutely wasn't AI generated. I don't think it's possible to get much insight from AI right now. It tends to the average and insight by definition is the opposite!
"Even if you hate OpenAI and ChatGPT for being responsible for the lack of job postings, I recommend you ally with them for now; learn to use ChatGPT before it's too late to keep your options open."
I think I will keep my options open without ChatGPT! I opt to do my own reading, thinking, and writing. I can't see any reason to trust what the broligarchy's algorithms spit out!
Just a few reasons to consider the human option:
1) ethical issues associated with tools based on intellectual theft of copyright-protected art and writing,
2) environmental costs of Gen AI,
3) horrific international labor practices of AI companies,
4) devaluation of human creativity, originality, critical thinking, and independent writing skills.
Those considerations are important. And should always be factored in the equation. Yet we use technology anyway because it's useful. The same applies to AI
But there are some use cases for ChatGPT, such as having AI auto-answer emails produced by AI and compressing AI slop back to the prompts that generated it. With any luck, ChatGPT will handle everything produced by ChatGPT so you don't have to.
I'm skeptical of points such as #6 and #10, and anything that claims there will be a demand for human made creative output. Like yes, there will be some demand, but it will be miniscule.
If we assume a metric of "maximal entertainment" then I don't think consumers will care if the creator is AI or human. All they will care about is what content at that moment in time provides the most entertainment. This applies to any metric you can apply to creative output. AI will be able to produce infinite on demand content catered specifically to it's audience. A human, even one enabled with AI tools, will unlikely be able to compete. Any audience for human made content will be from people trying to be "trendy". It will be far from the norm.
True but your failing to realize that who the human creator is, is a factor in the "maximum entertainment" you're trying to achieve. Things aren't funny by themselves but in a specific context that requires familiarity, subverted expectations and so on, which is often achieved by a specific comedian with a name. The same goes for content creation. AI can compete much better than us in certain aspects but not in all of them.
Not sure what your reply is supposed to mean. Just because there may still be some comedy not written by AI in 2 years time does not, in any way, imply that the non-AI jokes will be funnier. If you want me to be more explicit - In 2 years time all of the funniest jokes as voted for by audiences, will have been written by AI.
And will you go see an AI comedian? Or a human comedian whose name you know, who's manners you like, whose delivery and relationship with the audience you enjoy, whose tone, voice, accent, and presence you'd pay for? Really. You're thinking too small
Musings 2,4,7, and 9 are all essentially saying “Resistance is futile. You shall be assimilated.”
“8. Companies are choosing to reduce costs over increasing output because the sectors where generative AI is useful can't artificially increase demand in parallel to productivity. (Who needs more online content?)” is somewhat disingenuous. Companies choose to reduce costs to improve profits and shareholder value. Increasing output is a response to increased demand (or flooding the market to box out competitors).
The concept that won’t take your job is flawed. Consider a factory with two shifts. If coupling AI with a person increases productivity and demand stays the same, shift #2 is now in jeopardy. Alternatively, management will fire one person and give their job to an Ai assisted person who now has the capability to handle both.
May strike some as rather polyanish, however I suspect you are on the right track, broadly speaking. The only caveat I would offer is that I often meet investors and biz owners who positively salivate at not having any more human workers to "deal" with. So Capital may help drive the boat to mass unemployment, whether or not you work for yourself or someone else (not counting plumbers or ICE Workers LOL) I really enjoy the comments on here as well.
What possible justification can you have for point 10?
If AI can mimic even 90% of what we do creatively, let alone 99%, then it will soon mimic 100%
The main argument I have heard against this is actually worse than useless: People say that AI cannot be original when, in fact, it is people who cannot be original. Everything "original" that we create is based on our past experience. If I want LLM based AI to generate original output I only have to alter the weights of its neural network corresponding to its past experience slightly using a random number generator and there you have genuinely original output.
Basically you don't want to spend asimptotically more money to reach 99.9 then 99.99 and so on. It's cheap to approximate human labor but not to do it 100%. It's an extreme form of pareto distribution (the numbers 99 and 100 are arbitrary). Human skill allows us to get 100% because we've learned to do our stuff well. But AI doesn't work that way. As you try to get closer to 100% you need to spend exponentially more money to get the marginal extra benefit. For instance, automatic factory robots can do pretty much anything but he right person can do what they can't - but not because they're less skilled but because the money required to walk that last mile in ability *isn't worth it*.
Point 10 is about creativity not mimicing a human being. We are not trying to mimic some artist X, but creating an AI artist that is 100% as creative as artist X thus rendering X unemployable as an artist. There is no asymptotic performance limit inherent in this task (certainly none that you have identified).
Like #10, a bit like any human creation in the long long term acquiring finite crypto status relative to the infinitely replicable AI versions…its going to be an interesting one.
Yeah something like that. It's happened before with so many things. Human-made knives, brushes, hats, scissors, watches, suits - are the most expensive
Thanks Alberto - right on and very timely. PLEASE Amplify and clarify what you mean in #15.
I am stumped by it. What aother "timelines" and "AI Alignment" are your referring to? As a student of the impact of technologoy and innovatiopn on work, professions, and careers across industries over 4 decades, you nail it. Esp #12!
12. We should transition from “AI is eating the workforce” to “AI has created a new workforce” with safety nets and governmental help available to those who need it.
Thanks for your your good work - Please keep going. Matt,Sadinsky@prepintl.com
Thank you Matt! With timelines I refer to "what could have happened" scenarios (counterfactuals). With AI alignment I recommend starting on Wikipedia haha there's really a lot to read on that
In #15 - when you wrote, "There's a limited number of timelines where your job is the biggest of your worries:" What are you referring to? This is what I am confused about. The typical working life of adults 20's-70's is still a lifetime.
Sounds ominous "...and you may be dead if AI alignment is not solved." Are you refererring to significant events when relying on AI was inapproriate or misguided e.g. the failure of the weapon detection system @ Antioch HS and other failures? or somthing else. Thanks again.
I refer to the existential AI risk discourse. You can find plenty of info on Google! (Or ask ChatGPT haha)
The coolest thing about this post was that it really sounds like it wasn't generated by an AI - it's far more insightful than the average content on the internet that LLMs are trained on :)
Thank you Oliver. It absolutely wasn't AI generated. I don't think it's possible to get much insight from AI right now. It tends to the average and insight by definition is the opposite!
Well said! We will be able to identify the real Alberto for weeks to come.
This is a balanced and comprehensive article on impact of AI on the labour market.
Thank you Sylvia
"Even if you hate OpenAI and ChatGPT for being responsible for the lack of job postings, I recommend you ally with them for now; learn to use ChatGPT before it's too late to keep your options open."
I think I will keep my options open without ChatGPT! I opt to do my own reading, thinking, and writing. I can't see any reason to trust what the broligarchy's algorithms spit out!
Just a few reasons to consider the human option:
1) ethical issues associated with tools based on intellectual theft of copyright-protected art and writing,
2) environmental costs of Gen AI,
3) horrific international labor practices of AI companies,
4) devaluation of human creativity, originality, critical thinking, and independent writing skills.
Even the Pope has come out with concerns! "AI is ‘neither objective nor neutral" https://www.vaticannews.va/en/pope/news/2024-06/pope-g7-artifical-intelligence-objective-neutral.html
Those considerations are important. And should always be factored in the equation. Yet we use technology anyway because it's useful. The same applies to AI
But there are some use cases for ChatGPT, such as having AI auto-answer emails produced by AI and compressing AI slop back to the prompts that generated it. With any luck, ChatGPT will handle everything produced by ChatGPT so you don't have to.
I'm skeptical of points such as #6 and #10, and anything that claims there will be a demand for human made creative output. Like yes, there will be some demand, but it will be miniscule.
If we assume a metric of "maximal entertainment" then I don't think consumers will care if the creator is AI or human. All they will care about is what content at that moment in time provides the most entertainment. This applies to any metric you can apply to creative output. AI will be able to produce infinite on demand content catered specifically to it's audience. A human, even one enabled with AI tools, will unlikely be able to compete. Any audience for human made content will be from people trying to be "trendy". It will be far from the norm.
True but your failing to realize that who the human creator is, is a factor in the "maximum entertainment" you're trying to achieve. Things aren't funny by themselves but in a specific context that requires familiarity, subverted expectations and so on, which is often achieved by a specific comedian with a name. The same goes for content creation. AI can compete much better than us in certain aspects but not in all of them.
I fully expect most comedians to generate most of their material using AI within a maximum of 2 years.
Exactly. Most.
Not sure what your reply is supposed to mean. Just because there may still be some comedy not written by AI in 2 years time does not, in any way, imply that the non-AI jokes will be funnier. If you want me to be more explicit - In 2 years time all of the funniest jokes as voted for by audiences, will have been written by AI.
And will you go see an AI comedian? Or a human comedian whose name you know, who's manners you like, whose delivery and relationship with the audience you enjoy, whose tone, voice, accent, and presence you'd pay for? Really. You're thinking too small
Live? No. But AI generated on TV? Why not? And even "live" will be possible in future: Have you seen or heard about Abba Voyage?
https://w.wiki/CscK
Musings 2,4,7, and 9 are all essentially saying “Resistance is futile. You shall be assimilated.”
“8. Companies are choosing to reduce costs over increasing output because the sectors where generative AI is useful can't artificially increase demand in parallel to productivity. (Who needs more online content?)” is somewhat disingenuous. Companies choose to reduce costs to improve profits and shareholder value. Increasing output is a response to increased demand (or flooding the market to box out competitors).
What I'm saying is that you can't increase demand artificially like you can productivity
The concept that won’t take your job is flawed. Consider a factory with two shifts. If coupling AI with a person increases productivity and demand stays the same, shift #2 is now in jeopardy. Alternatively, management will fire one person and give their job to an Ai assisted person who now has the capability to handle both.
https://deepseekv3.tech/
May strike some as rather polyanish, however I suspect you are on the right track, broadly speaking. The only caveat I would offer is that I often meet investors and biz owners who positively salivate at not having any more human workers to "deal" with. So Capital may help drive the boat to mass unemployment, whether or not you work for yourself or someone else (not counting plumbers or ICE Workers LOL) I really enjoy the comments on here as well.
What possible justification can you have for point 10?
If AI can mimic even 90% of what we do creatively, let alone 99%, then it will soon mimic 100%
The main argument I have heard against this is actually worse than useless: People say that AI cannot be original when, in fact, it is people who cannot be original. Everything "original" that we create is based on our past experience. If I want LLM based AI to generate original output I only have to alter the weights of its neural network corresponding to its past experience slightly using a random number generator and there you have genuinely original output.
Basically you don't want to spend asimptotically more money to reach 99.9 then 99.99 and so on. It's cheap to approximate human labor but not to do it 100%. It's an extreme form of pareto distribution (the numbers 99 and 100 are arbitrary). Human skill allows us to get 100% because we've learned to do our stuff well. But AI doesn't work that way. As you try to get closer to 100% you need to spend exponentially more money to get the marginal extra benefit. For instance, automatic factory robots can do pretty much anything but he right person can do what they can't - but not because they're less skilled but because the money required to walk that last mile in ability *isn't worth it*.
Point 10 is about creativity not mimicing a human being. We are not trying to mimic some artist X, but creating an AI artist that is 100% as creative as artist X thus rendering X unemployable as an artist. There is no asymptotic performance limit inherent in this task (certainly none that you have identified).
I'm talking about creativity and craft
Like #10, a bit like any human creation in the long long term acquiring finite crypto status relative to the infinitely replicable AI versions…its going to be an interesting one.
Yeah something like that. It's happened before with so many things. Human-made knives, brushes, hats, scissors, watches, suits - are the most expensive