Ian Bremmer and Mustafa Suleyman on world power relationships; Freddie deBoer is not impressed with the latest AI; Tyler Cowen on limits to growth; The Zvi disagrees with Tyler
How do we expect AI to improve the world economy? Having it replace certain jobs is an incremental change - technology has long been destroying old jobs and creating new jobs. How will AI replace jobs in a far superior way?
Take software programming. If AI is used to build software, do humans need to understand what AI coded? The big leap would be for humans to turn software coding and testing over to the computers. That could be a huge productivity jump. But at what risk? Who validates the AI? Who is liable for what AI codes?
There is a great looming problem with AI regulation. It is that modern America has embraced a regulatory culture that it is ok to do bad things because (1) you might get away with it and (2) if caught the punishment makes the bad thing worth it. This history informs me that companies will use AI in unethical ways and this will lead to large system / social failures.
I sense the AI proponents are ignoring the cost of these failures in their projections. I wonder if the AI proponents even imagine that AI can produce great failures. Such is the hubris of technocrats. My forecast is we will see no great improvement in social welfare but we will see increasing inequality as ever more gains are privatized and losses socialized.
The industrial revolution greatly increased the population, which is itself an important factor of production. Likely the increase in population did more for gdp than the industrial revolution did otherwise.
I don't see AI adding to the population, so I think it's contribution to gdp will be much smaller than for the IR-- even if you think it's game changing technology.
A lot of how LLMs shake out will be determined by whether the courts take the "Fair Learning" perspective from Mark Lemley (https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3528447) or the perspective preferred by the big IP holders, which is that training is infringement. Even if Hollywood wins, there's still a lot that can be done with the technology.
One way to think about the Chips Act, the preparation to defend Taiwan, and the proliferation of AI audit bills in state legislatures is that the environment is going to modestly slow AI projects in the places that we expect AI projects. But if you squeeze a balloon ...
“AI’s trajectory will be largely determined by the decisions of a handful of private businesses...” Isn’t this contrary to the AI Doom - that AI does it’s own thing beyond the reach of Humans? And it is an oddity that it should be considered more of a threat than PI - political intelligence that infects Governments and their stooges and the serial disasters it inflicts on Humanity. If AI will eradicate PI - where do I sign?
The far out Doom is the Superintelligence eliminating all of humanity because we can not compete with it and we lose the evolutionary race in total. The somewhat more near doom is that people do bad things with the new tools on the way to Superintelligence similar to the other bad things people have done throughout history. At least that is my reading of the multiple future doom scenarios.
I’ve filed it under - New Ice Age, Population Explosion, famine and food wars, mass unemployment because of robots, Climate Change, World population destroyed by SARS CoV 1… correction SARS CoV 2… correction SARS CoV 3 (soon to be announced).
Don’t forget. Computers, even AI, run on electricity. Soon there won’t be any thanks to one of the previous DOOMS.
The early anecdotes yield a pessimistic view. For while it is impressive AI can compose essays and mimic personalities, other tests show AI to be very dumb - meaning garbage in = garbage out. What this indicates is AI is just a tool and the tool will magnify the intelligence and ignorance of the user. Given where we are in modern society and the amount of foolishness that is promoted as "The Science", one has cause to be pessimistic about how AI will be used and the influence it will have.
I agree that AI , like ChatGPT, is just a tool and it's about how you use it. But most successful technologies in their early stages of development--PCs, the Internet, etc., start off seemingly to have minimal uses and end up becoming essential. Do you disagree this is not the case for some AIs?
Never underestimate the ability of humans to create a need for a product they want to sell. But right now I see a pressing need for AI copy editors. So does Microsoft:
"Microsoft removed a set of bizarre travel articles made with 'algorithmic techniques.' But it won't blame AI."
I was even thinking about selling the September 24 calls around $500-600/share to fund some of the put buys. I haven't pulled the trigger since it could easily go to $1500/share in this silly environment.
They are the best, their CUDA investment 8 years ago is giving them a top performance monopoly. Selling all they can make thru Q3, 2024. I’m thinking of buying…
How do we expect AI to improve the world economy? Having it replace certain jobs is an incremental change - technology has long been destroying old jobs and creating new jobs. How will AI replace jobs in a far superior way?
Take software programming. If AI is used to build software, do humans need to understand what AI coded? The big leap would be for humans to turn software coding and testing over to the computers. That could be a huge productivity jump. But at what risk? Who validates the AI? Who is liable for what AI codes?
There is a great looming problem with AI regulation. It is that modern America has embraced a regulatory culture that it is ok to do bad things because (1) you might get away with it and (2) if caught the punishment makes the bad thing worth it. This history informs me that companies will use AI in unethical ways and this will lead to large system / social failures.
I sense the AI proponents are ignoring the cost of these failures in their projections. I wonder if the AI proponents even imagine that AI can produce great failures. Such is the hubris of technocrats. My forecast is we will see no great improvement in social welfare but we will see increasing inequality as ever more gains are privatized and losses socialized.
The industrial revolution greatly increased the population, which is itself an important factor of production. Likely the increase in population did more for gdp than the industrial revolution did otherwise.
I don't see AI adding to the population, so I think it's contribution to gdp will be much smaller than for the IR-- even if you think it's game changing technology.
A lot of how LLMs shake out will be determined by whether the courts take the "Fair Learning" perspective from Mark Lemley (https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3528447) or the perspective preferred by the big IP holders, which is that training is infringement. Even if Hollywood wins, there's still a lot that can be done with the technology.
One way to think about the Chips Act, the preparation to defend Taiwan, and the proliferation of AI audit bills in state legislatures is that the environment is going to modestly slow AI projects in the places that we expect AI projects. But if you squeeze a balloon ...
“AI’s trajectory will be largely determined by the decisions of a handful of private businesses...” Isn’t this contrary to the AI Doom - that AI does it’s own thing beyond the reach of Humans? And it is an oddity that it should be considered more of a threat than PI - political intelligence that infects Governments and their stooges and the serial disasters it inflicts on Humanity. If AI will eradicate PI - where do I sign?
Why would it eliminate "PI"? The concern is that it will amplify it.
Because we shall all be ruled by machines. Isn’t that the Doom?
The far out Doom is the Superintelligence eliminating all of humanity because we can not compete with it and we lose the evolutionary race in total. The somewhat more near doom is that people do bad things with the new tools on the way to Superintelligence similar to the other bad things people have done throughout history. At least that is my reading of the multiple future doom scenarios.
I’ve filed it under - New Ice Age, Population Explosion, famine and food wars, mass unemployment because of robots, Climate Change, World population destroyed by SARS CoV 1… correction SARS CoV 2… correction SARS CoV 3 (soon to be announced).
Don’t forget. Computers, even AI, run on electricity. Soon there won’t be any thanks to one of the previous DOOMS.
It also sounds like Bremmer and Suleyman are pessimistic on AI's ability to improve the world. It seems way to early to be pessimistic on LLMs etc.
The early anecdotes yield a pessimistic view. For while it is impressive AI can compose essays and mimic personalities, other tests show AI to be very dumb - meaning garbage in = garbage out. What this indicates is AI is just a tool and the tool will magnify the intelligence and ignorance of the user. Given where we are in modern society and the amount of foolishness that is promoted as "The Science", one has cause to be pessimistic about how AI will be used and the influence it will have.
I agree that AI , like ChatGPT, is just a tool and it's about how you use it. But most successful technologies in their early stages of development--PCs, the Internet, etc., start off seemingly to have minimal uses and end up becoming essential. Do you disagree this is not the case for some AIs?
Never underestimate the ability of humans to create a need for a product they want to sell. But right now I see a pressing need for AI copy editors. So does Microsoft:
"Microsoft removed a set of bizarre travel articles made with 'algorithmic techniques.' But it won't blame AI."
https://www.businessinsider.com/microsoft-removes-embarrassing-offensive-ai-assisted-travel-articles-2023-8
I was thinking about buying puts on NVDA ever since it crossed $400/share.
I was even thinking about selling the September 24 calls around $500-600/share to fund some of the put buys. I haven't pulled the trigger since it could easily go to $1500/share in this silly environment.
They are the best, their CUDA investment 8 years ago is giving them a top performance monopoly. Selling all they can make thru Q3, 2024. I’m thinking of buying…