Jason Furman on regulatory principles; Mathis Bitton and Jack Sadler against screens in schools; Benedict Evans on AI doomers; Matt Yglesias on employment doom
On the Yglesias article, I think the secretaries and office managers are precisely people who will not lose their jobs to AI. They do a lot of random different things (so maybe LLMs can do some of them but not all of them, maybe LLMs make them more productive and they start doing other things like bank tellers) and have already been disrupted by previous rounds of technological innovation.
Hear, hear for hostility to the term “AI.” I’ve worked in machine learning for almost 20 years, and I still don’t understand what these deep learning models are supposed to be used for. VisiCalc was directly useful, but these models can, um, generate generic digital content?
Seems like a lot of folks hear the term “Artificial Intelligence” and start reasoning from there, imagining Skynet and whatnot, regardless of what these models actually are. What would’ve happened if the models had a different, more boring name like “serial delinearized regression” (because that’s what the models are) or something? Would any of these commentators even be aware of them?
Quotes from AC Clarke and AN Whitehead are ringing in my ears. Lessons from the arrival of the internet 1990-2020:
1. Fears of disintermediation became lots of new intermediaries (e.g., Uber and food delivery via apps) as unit costs fell.
2. More complementary goods and substitutions as Apple Maps took over from Mapquest, which had taken over from Rand McNally.
3. Slow adoption by organizations for all the usual resistance-to-innovation reasons, so innovation to happen amongst the unserved if one believes Clay Christensen, possibly by new entrants.
Other random thoughts: I believe the majority of occupations require certification, so the guilds are unlikely to allow AI to be unleashed. Real estate agents are still with us, travel agents less so but not zero -- some people prefer self-serve, others someone else to blame. What is the opportunity cost for the most productive to "supervise" AI versus hiring someone else to do so? Major failures will be excuses for innovation-slowing regulation (e.g., 9-11 and TSA), whether failures were due to human error or "unsupervised" AI. Sky yet to fall despite centuries of doomsayers.
More broadly, paying people not to work and the tax wedge are likely far more destructive of jobs. The end of illegal immigration will improve low-skilled workers' bargaining position across the West, and increase the need for and use of labor-substituting capital goods that need to be developed, manufactured, sold and serviced (e.g., iRobot lawnmowers). The good news will include more pressure on the K-12 Ed system to produce workers with skills -- I just saw my first billboard for a Master's in AI. More of a premium on people and trade skills?
Re: Depressions, innovations, and labor displacement.
There is a set of television programs called the BBC historic farm series that are an absolute treasure. There are several on Amazon Prime right now but can also be watched on YouTube. There is one on the Victorian Farm set in the 1880s and another on the Edwardian Farm set in the 1900 to 1914 period. You can watch them back to back to see how little had changed between the two eras they present. However, on the Victorian farm there are numerous inventions that are being used and became common that were not widely available 30 and 40 years earlier. When one looks up the Long Depression (1873-1879) the parallels to the great depression and the amount of farm labor displacement that may have been possible from horse drawn reaper binders and better wheat threshing machines is interesting. The government of the Grant administration also screwed up monetary policy and google claims, "The Panic of 1873 was triggered by overexpansion of railroads and speculative investments, which outstripped consumer demand." https://en.wikipedia.org/wiki/BBC_historic_farm_series
After the Civil War, ag production surged due to increased technological efficiency, including railroads, and new farm entrants -- largely immigrants. This brought down farm prices and squeezed farm incomes. There really wasn't much in the way of monetary policy at the time other than gold flows but the government was dedicated to returning the "Greenback" to parity, which happened in 1879. So monetary policy must have been tight. I think there was a big influx of silver, which made farmers and WJ Bryan want a silver standard.
Yeah I probably should have stated that more neutrally with regard to money supply issues, and prevailing wisdom of the time might have been a tight monetary policy no matter what.
It took an entire day. I dealt with several people all of which seemed to involve some kind of negotiation or paperwork that was unnecessary in the grand scheme of things. The salesperson "on my side" didn't do anything for me other then try to keep me from arguing directly with the guy trying to rip me off on my trade in. I had to waste time driving over to a rival dealership down the road to get a better quote on my trade in when we could have just been civil with each other to start.
There were a lot of people in this dealership doing "bullshit jobs". While an LLM could probably regurgitate the same answers as they did face to face, the face to face is obviously the point. Dealing with an AI would just have agitated me further.
At the end, sensing I was going to give him a very bad review with the car company, the owner tried to say that the review would only affect the poor guy "on my side" during the whole thing and not the guy that tried to screw me on my trade in, so I shouldn't "take it out on him."
And this does't even get into the fact that car dealers probably aren't necessary middle men at all but are protected by law.
Jobs that seem pointless may in fact be the most robust against AI.
A friend of mine used to show up for such purchases with a cashier’s check. Which she would triumphantly pull out of her purse after hours of haggling. This prompted comments, I recall, ranging from “You had it all along?” to “Your husband must be a very patient man”. I wonder at what point you could bring it out, and whether it would have any effect, if you hadn’t her steely resolve.
World War II brought the US out of the Great Depression in spite of its colossal waste in material and lives. In 1939 the unemployment rate still lingered around 18%. (To me, the big question about the Depression is why the economy re-collapsed in 1937.)
I have always wondered if maybe the New Deal would have done the job if it had been like a hundred times bigger -- the size of the war effort. Suppose that in 1932 the government had contracted with manufacturers to quadruple production and just dump everything in the ocean, which is pretty much what happened in the war. Oh, and printed money to cover the resulting fiscal deficit. Would this have ended the Depression with a similar inflationary cost? I don't know why not.
Friedman and Schwartz cover this in their Monetary History of the United States, 1867-1960. The Fed, looking at the large excess reserves the banks had accumulated by 1936 or so, thought, "OMG, we might have a bout of inflation when they start lending." So the Fed doubled reserve requirements. The banks responded by reducing lending and calling in loans. What the Fed didn't take account of is that the bankers learned from the early 1930s and didn't want to be caught again with too few reserves. (I'm going by memory here: my copy of the Friedman/Schwartz book is at my office.)
"Maybe relatively few people will “sign up” for AI; instead, it will just become part of the background of how things work." That's what's happening in my home. My RN wife has installed Claude on her phone and asks it questions about various things. It's a much easier process than web search. She hasn't noticed any medical errors so far. I have Claude on my desktop and Perplexity on my phone. I ask them questions occasionally, and they usually provide useful information.
But in areas of physical experience, they can be laughably inept. I asked Claude how to execute a martellato while playing handbells. The answer was analogous to a child's first drawing of her mother. It's actually rather a difficult thing to describe in words. A human would have responded by grabbing a handbell and demonstrating. In the absence of a physical handbell, a human might have found a YouTube video demonstration. But Claude, at least at that point, had only words to use, and words failed.
"But if you were to go back in history when writing was invented, could you not make similar arguments about what people will lose when they no longer learn to recite epic poetry from memory?"
No doubt something was lost when many fewer were able to tell epic poems and stories. Likewise much was gained from reading and writing.
While handwriting surely has some benefit as brain exercise, it's less clear to me why, or even if, that particular exercise is more valuable than any other.
"...could you not make similar arguments about what people will lose when they no longer learn to recite epic poetry from memory?"
I imagine a poem in those circumstances cannot but change over the generations and possibly become so different as to be unrecognizable. And no one would ever know because it wasn't carved in stone somewhere.
"There is an old Egyptian tale of Theuth, the inventor of writing, showing his invention to the god Thamus, who told him that he would only spoil men's memories and take away their understandings. From this tale, of which young Athens will probably make fun, may be gathered the lesson that writing is inferior to speech. For it is like a picture, which can give no answer to a question, and has only a deceitful likeness of a living creature. It has no power of adaptation, but uses the same words for all. It is not a legitimate son of knowledge, but a bastard, and when an attack is made upon this bastard neither parent nor anyone else is there to defend it. The husbandman will not seriously incline to sow his seed in such a hot-bed or garden of Adonis; he will rather sow in the natural soil of the human soul which has depth of earth; and he will anticipate the inner growth of the mind, by writing only, if at all, as a remedy against old age. The natural process will be far nobler, and will bring forth fruit in the minds of others as well as in his own."
Cicero, pseudo-Cicero (adaptations not actually written by him but believed incorrectly to be him), and then Thomas Aquinas tried to mitigate some of these downsides to the adoption of writing by advocating a discipline of contemplating images to aid memory. But I guess another way of mitigating the negative aspects of mass literacy used by most historical societies is just limiting it to a small class of specialists, much like how we mitigate the incidence of black lung by only have some people be coal miners. We may be returning to such a social arrangement inadvertently.
On the Yglesias article, I think the secretaries and office managers are precisely people who will not lose their jobs to AI. They do a lot of random different things (so maybe LLMs can do some of them but not all of them, maybe LLMs make them more productive and they start doing other things like bank tellers) and have already been disrupted by previous rounds of technological innovation.
Hear, hear for hostility to the term “AI.” I’ve worked in machine learning for almost 20 years, and I still don’t understand what these deep learning models are supposed to be used for. VisiCalc was directly useful, but these models can, um, generate generic digital content?
Seems like a lot of folks hear the term “Artificial Intelligence” and start reasoning from there, imagining Skynet and whatnot, regardless of what these models actually are. What would’ve happened if the models had a different, more boring name like “serial delinearized regression” (because that’s what the models are) or something? Would any of these commentators even be aware of them?
Quotes from AC Clarke and AN Whitehead are ringing in my ears. Lessons from the arrival of the internet 1990-2020:
1. Fears of disintermediation became lots of new intermediaries (e.g., Uber and food delivery via apps) as unit costs fell.
2. More complementary goods and substitutions as Apple Maps took over from Mapquest, which had taken over from Rand McNally.
3. Slow adoption by organizations for all the usual resistance-to-innovation reasons, so innovation to happen amongst the unserved if one believes Clay Christensen, possibly by new entrants.
Other random thoughts: I believe the majority of occupations require certification, so the guilds are unlikely to allow AI to be unleashed. Real estate agents are still with us, travel agents less so but not zero -- some people prefer self-serve, others someone else to blame. What is the opportunity cost for the most productive to "supervise" AI versus hiring someone else to do so? Major failures will be excuses for innovation-slowing regulation (e.g., 9-11 and TSA), whether failures were due to human error or "unsupervised" AI. Sky yet to fall despite centuries of doomsayers.
More broadly, paying people not to work and the tax wedge are likely far more destructive of jobs. The end of illegal immigration will improve low-skilled workers' bargaining position across the West, and increase the need for and use of labor-substituting capital goods that need to be developed, manufactured, sold and serviced (e.g., iRobot lawnmowers). The good news will include more pressure on the K-12 Ed system to produce workers with skills -- I just saw my first billboard for a Master's in AI. More of a premium on people and trade skills?
Re: Depressions, innovations, and labor displacement.
There is a set of television programs called the BBC historic farm series that are an absolute treasure. There are several on Amazon Prime right now but can also be watched on YouTube. There is one on the Victorian Farm set in the 1880s and another on the Edwardian Farm set in the 1900 to 1914 period. You can watch them back to back to see how little had changed between the two eras they present. However, on the Victorian farm there are numerous inventions that are being used and became common that were not widely available 30 and 40 years earlier. When one looks up the Long Depression (1873-1879) the parallels to the great depression and the amount of farm labor displacement that may have been possible from horse drawn reaper binders and better wheat threshing machines is interesting. The government of the Grant administration also screwed up monetary policy and google claims, "The Panic of 1873 was triggered by overexpansion of railroads and speculative investments, which outstripped consumer demand." https://en.wikipedia.org/wiki/BBC_historic_farm_series
After the Civil War, ag production surged due to increased technological efficiency, including railroads, and new farm entrants -- largely immigrants. This brought down farm prices and squeezed farm incomes. There really wasn't much in the way of monetary policy at the time other than gold flows but the government was dedicated to returning the "Greenback" to parity, which happened in 1879. So monetary policy must have been tight. I think there was a big influx of silver, which made farmers and WJ Bryan want a silver standard.
Yeah I probably should have stated that more neutrally with regard to money supply issues, and prevailing wisdom of the time might have been a tight monetary policy no matter what.
I bought a car on Friday.
It took an entire day. I dealt with several people all of which seemed to involve some kind of negotiation or paperwork that was unnecessary in the grand scheme of things. The salesperson "on my side" didn't do anything for me other then try to keep me from arguing directly with the guy trying to rip me off on my trade in. I had to waste time driving over to a rival dealership down the road to get a better quote on my trade in when we could have just been civil with each other to start.
There were a lot of people in this dealership doing "bullshit jobs". While an LLM could probably regurgitate the same answers as they did face to face, the face to face is obviously the point. Dealing with an AI would just have agitated me further.
At the end, sensing I was going to give him a very bad review with the car company, the owner tried to say that the review would only affect the poor guy "on my side" during the whole thing and not the guy that tried to screw me on my trade in, so I shouldn't "take it out on him."
And this does't even get into the fact that car dealers probably aren't necessary middle men at all but are protected by law.
Jobs that seem pointless may in fact be the most robust against AI.
A friend of mine used to show up for such purchases with a cashier’s check. Which she would triumphantly pull out of her purse after hours of haggling. This prompted comments, I recall, ranging from “You had it all along?” to “Your husband must be a very patient man”. I wonder at what point you could bring it out, and whether it would have any effect, if you hadn’t her steely resolve.
World War II brought the US out of the Great Depression in spite of its colossal waste in material and lives. In 1939 the unemployment rate still lingered around 18%. (To me, the big question about the Depression is why the economy re-collapsed in 1937.)
I have always wondered if maybe the New Deal would have done the job if it had been like a hundred times bigger -- the size of the war effort. Suppose that in 1932 the government had contracted with manufacturers to quadruple production and just dump everything in the ocean, which is pretty much what happened in the war. Oh, and printed money to cover the resulting fiscal deficit. Would this have ended the Depression with a similar inflationary cost? I don't know why not.
Friedman and Schwartz cover this in their Monetary History of the United States, 1867-1960. The Fed, looking at the large excess reserves the banks had accumulated by 1936 or so, thought, "OMG, we might have a bout of inflation when they start lending." So the Fed doubled reserve requirements. The banks responded by reducing lending and calling in loans. What the Fed didn't take account of is that the bankers learned from the early 1930s and didn't want to be caught again with too few reserves. (I'm going by memory here: my copy of the Friedman/Schwartz book is at my office.)
Without the amazing innovation that was driven by the necessities of war, your idea would destroy the country.
"Maybe relatively few people will “sign up” for AI; instead, it will just become part of the background of how things work." That's what's happening in my home. My RN wife has installed Claude on her phone and asks it questions about various things. It's a much easier process than web search. She hasn't noticed any medical errors so far. I have Claude on my desktop and Perplexity on my phone. I ask them questions occasionally, and they usually provide useful information.
But in areas of physical experience, they can be laughably inept. I asked Claude how to execute a martellato while playing handbells. The answer was analogous to a child's first drawing of her mother. It's actually rather a difficult thing to describe in words. A human would have responded by grabbing a handbell and demonstrating. In the absence of a physical handbell, a human might have found a YouTube video demonstration. But Claude, at least at that point, had only words to use, and words failed.
> AI has yet to have its “Visicalc moment,” where a large segment of users looks and realizes: I must have this.
Wasn’t this ChatGPT? It quickly went from 0 to billions of revenue, which is no small matter
"But if you were to go back in history when writing was invented, could you not make similar arguments about what people will lose when they no longer learn to recite epic poetry from memory?"
No doubt something was lost when many fewer were able to tell epic poems and stories. Likewise much was gained from reading and writing.
While handwriting surely has some benefit as brain exercise, it's less clear to me why, or even if, that particular exercise is more valuable than any other.
"...could you not make similar arguments about what people will lose when they no longer learn to recite epic poetry from memory?"
I imagine a poem in those circumstances cannot but change over the generations and possibly become so different as to be unrecognizable. And no one would ever know because it wasn't carved in stone somewhere.
From Plato's dialogue with Phaedrus:
"There is an old Egyptian tale of Theuth, the inventor of writing, showing his invention to the god Thamus, who told him that he would only spoil men's memories and take away their understandings. From this tale, of which young Athens will probably make fun, may be gathered the lesson that writing is inferior to speech. For it is like a picture, which can give no answer to a question, and has only a deceitful likeness of a living creature. It has no power of adaptation, but uses the same words for all. It is not a legitimate son of knowledge, but a bastard, and when an attack is made upon this bastard neither parent nor anyone else is there to defend it. The husbandman will not seriously incline to sow his seed in such a hot-bed or garden of Adonis; he will rather sow in the natural soil of the human soul which has depth of earth; and he will anticipate the inner growth of the mind, by writing only, if at all, as a remedy against old age. The natural process will be far nobler, and will bring forth fruit in the minds of others as well as in his own."
What a wretched species we would remain if that advice had taken hold.
Cicero, pseudo-Cicero (adaptations not actually written by him but believed incorrectly to be him), and then Thomas Aquinas tried to mitigate some of these downsides to the adoption of writing by advocating a discipline of contemplating images to aid memory. But I guess another way of mitigating the negative aspects of mass literacy used by most historical societies is just limiting it to a small class of specialists, much like how we mitigate the incidence of black lung by only have some people be coal miners. We may be returning to such a social arrangement inadvertently.
O brave new world!