Since the very first days of the internet, my nickname has been 'Luddite'. It was meant to be ironic, but I'm wondering if 'prescient' might have been a better label. I've also noticed my spell/grammar checker going way beyond its original remit and trying to re-write everything. It's particularly fond of applying a liberal sprinkling of commas and reducing text to a colourless wall of corporate speak - not helpful when I'm trying to write a piece of gripping fiction.
This is so timely, Dr. Jo, as I just finished binging the first season of Pluribus AND just listened to a podcast all about the Luddites! So disappointing to see that nothing had really changed since the Luddites were around -- there are now entire countries (U, cough, cough, S) that are completely owned by wealthy corporations whose primary goal is to chew up and spit out laborers while paying them as little as possible! After all Daddy CEO needs a new yacht!
It is tempting to call out what you see as genAI slop but when you get it wrong? Several authors here are using these tools as translators, broadening their reach to an English-centric audience. I've no way to confirm that works produced this way end up appearing like slop, so have stopped calling most out.
Not as resistive a tactic perhaps, I simply mute sloppy-seeming authors instead.
Yes. It _is_ increasingly difficult to tease out whether someone is generating content using AI, or simply using a tool to broaden their reach. And as I've noted before ( https://drjo.substack.com/i/179102888/its-about-quality ) what really counts is quality; AI will likely eventually surpass us there.
I think we're mostly on the same page. We can criticise bad content;* we can also take appropriate actions; and we can at least try to resist the imposition of awful decisions by powerful people. The alternative is dreadful.
Dr Jo.
=======
*And if we incorrectly impute that someone used AI, they can call us out in turn, and perhaps we've learnt something.
Weather prediction nowadays quite often begins with a reasonable weather model, the current weather conditions, and then executes multiple runs of the model and averages.
You can sometimes see these families of runs when it comes to predictions of cyclones. In the United States, these families of models are sometimes augmented by Sharpie Markers, but that is beside the point.
My proposal is this: create a simplified model of individual human behaviour, then take one of those big supercomputer clusters and run populations of these models to predict human behaviour in the aggregate. Then you can try out different ways of affecting the outcome.
Another possibility with the reasonable model of human behaviour is to use genetic learning algorithms on collections of populations. This can be used to derive socioeconomic rules that can help find optimal solutions to social problems.
Enticing thoughts. That was the appeal of Asimov's _Foundation_ novels. Problem is that people are even worse than the weather, where it's extraordinarily difficult to have a stab at conditions more than a few months in the future, 'cos chaos (and the Quantum Weather Butterfly :)
Your idea is unfortunately also refuted by random things like Trump and labubus and '67' (don't ask).
What we can do (rather than trying to *predict*) is something similar where we try at least to get the basics right, in terms of game theory, heuristics that work, and trying to be fair. Even if we're basically f-cked, I'd rather go down espousing things like good education and giving everyone a fair go, as opposed to the current tendency towards greed, fascism, xenophobia and armed conflict.
Actually, no: my idea is not refuted. Keep in mind that the weather is modeled by nonlinear partial differential equations. The Butterfly example is why the weather is so hard, not human society.
If anything, human society is easier because people act rationally sometimes, and the rest of the time they act emotionally but with decades of psychological understanding behind it. We are less random than the weather.
Your example of "random things like Trump" - well, I shall just give an example dear to your heart - that of an emergency room. You have probably come across many examples of freak accidents that are one of a kind. And yet, the majority of cases are dealt with through practice and experience.
This drives home my point - humans have commonalities we can expect, predict and model. What is complicated are the interactions. The weather is like that. The weather is made up of molecules of nitrogen, oxygen, water and so on. You run multiple simulations to average out the random effects.
Game theory is a simple first-order approximation. With a decent model of human behaviour, we can do much, much better by running multiple simulations. What I am considering is a sophisticated version of a Monte Carlo method.
As far as espousing solutions, that is a fundamental problem with politics today. Everybody is espousing solutions. The thing about human nature is that due to the complex interactions, there are so many unintended consequences that you cannot think of. A "good education" - does this include school uniforms? A fair go - what is the definition of fair?
I pose this question to you. If "greed, fascism, xenophobia and armed conflict" is so much worse, why do people choose that? A certain amount of greed leads to successful businesses. Fascism is popular nowadays because democracy has failed the working man. Xenophobia is bred in the bone - curing genetic diseases is tough enough. And as far as armed conflict, well every day I am grateful that the brave soldiers of Ukraine are standing up to Putin's minions so that I don't have to do it in the streets of Boston. And yet, the soldiers they are fighting also believe their cause if just. Again I say, it would help to run some serious simulations that try to answer the question of why people go for these solutions.
My worry is that if you go beyond the platitudes and actually try to put into place the things you espouse, the cure would be as bad as the disease. Think of the people who stuck with diet to cure ulcers or targeting Alzheimer's plaques. Many of them meant well - not all were charlatans.
Before you go espousing things, I suggest you run some simulations.
By the way, I am currently working on a book about meta-ethics. Being a computer techie with a bachelor's in phyiscs, I am doing stuff like modeling the is-ought distinstion as a stress-energy tensor. There is no such thing as a pure virtue or vice. Aristotle was right - there is a Golden Mean.
And having worked in Machine Learning, there is a fundamental theorem in Ensemble Learning (think: society of weak learners) that shows that if your weak learners operate at better than chance (right 51% of the time), then the ensemble will Boost the results of the weak learners. Consider a dictator (Xi Jinping, for example). A dictator is the Ensemble method that combines the efforts of the individuals. If his direction is better than chance, the society functions more efficiently than a herd-of-cats democracy, because democracy is inefficient. On the other hand, if the dictator is worse than chance (Vladimir Putin, for example) then society goes into the ditch very fast. That is why fascism is chosen so often, but democracy wins in the end. If American Democracy is sound, it will be self-correcting after this administration.
Since the very first days of the internet, my nickname has been 'Luddite'. It was meant to be ironic, but I'm wondering if 'prescient' might have been a better label. I've also noticed my spell/grammar checker going way beyond its original remit and trying to re-write everything. It's particularly fond of applying a liberal sprinkling of commas and reducing text to a colourless wall of corporate speak - not helpful when I'm trying to write a piece of gripping fiction.
:) Another good reason to abandon Windows and its soulless apps for Linux.
This is so timely, Dr. Jo, as I just finished binging the first season of Pluribus AND just listened to a podcast all about the Luddites! So disappointing to see that nothing had really changed since the Luddites were around -- there are now entire countries (U, cough, cough, S) that are completely owned by wealthy corporations whose primary goal is to chew up and spit out laborers while paying them as little as possible! After all Daddy CEO needs a new yacht!
It is tempting to call out what you see as genAI slop but when you get it wrong? Several authors here are using these tools as translators, broadening their reach to an English-centric audience. I've no way to confirm that works produced this way end up appearing like slop, so have stopped calling most out.
Not as resistive a tactic perhaps, I simply mute sloppy-seeming authors instead.
Yes. It _is_ increasingly difficult to tease out whether someone is generating content using AI, or simply using a tool to broaden their reach. And as I've noted before ( https://drjo.substack.com/i/179102888/its-about-quality ) what really counts is quality; AI will likely eventually surpass us there.
I think we're mostly on the same page. We can criticise bad content;* we can also take appropriate actions; and we can at least try to resist the imposition of awful decisions by powerful people. The alternative is dreadful.
Dr Jo.
=======
*And if we incorrectly impute that someone used AI, they can call us out in turn, and perhaps we've learnt something.
Well, here is my proposal.
Weather prediction nowadays quite often begins with a reasonable weather model, the current weather conditions, and then executes multiple runs of the model and averages.
You can sometimes see these families of runs when it comes to predictions of cyclones. In the United States, these families of models are sometimes augmented by Sharpie Markers, but that is beside the point.
My proposal is this: create a simplified model of individual human behaviour, then take one of those big supercomputer clusters and run populations of these models to predict human behaviour in the aggregate. Then you can try out different ways of affecting the outcome.
Another possibility with the reasonable model of human behaviour is to use genetic learning algorithms on collections of populations. This can be used to derive socioeconomic rules that can help find optimal solutions to social problems.
Enticing thoughts. That was the appeal of Asimov's _Foundation_ novels. Problem is that people are even worse than the weather, where it's extraordinarily difficult to have a stab at conditions more than a few months in the future, 'cos chaos (and the Quantum Weather Butterfly :)
Your idea is unfortunately also refuted by random things like Trump and labubus and '67' (don't ask).
What we can do (rather than trying to *predict*) is something similar where we try at least to get the basics right, in terms of game theory, heuristics that work, and trying to be fair. Even if we're basically f-cked, I'd rather go down espousing things like good education and giving everyone a fair go, as opposed to the current tendency towards greed, fascism, xenophobia and armed conflict.
At least, that's how I see it, Dr Jo.
Actually, no: my idea is not refuted. Keep in mind that the weather is modeled by nonlinear partial differential equations. The Butterfly example is why the weather is so hard, not human society.
If anything, human society is easier because people act rationally sometimes, and the rest of the time they act emotionally but with decades of psychological understanding behind it. We are less random than the weather.
Your example of "random things like Trump" - well, I shall just give an example dear to your heart - that of an emergency room. You have probably come across many examples of freak accidents that are one of a kind. And yet, the majority of cases are dealt with through practice and experience.
This drives home my point - humans have commonalities we can expect, predict and model. What is complicated are the interactions. The weather is like that. The weather is made up of molecules of nitrogen, oxygen, water and so on. You run multiple simulations to average out the random effects.
Game theory is a simple first-order approximation. With a decent model of human behaviour, we can do much, much better by running multiple simulations. What I am considering is a sophisticated version of a Monte Carlo method.
As far as espousing solutions, that is a fundamental problem with politics today. Everybody is espousing solutions. The thing about human nature is that due to the complex interactions, there are so many unintended consequences that you cannot think of. A "good education" - does this include school uniforms? A fair go - what is the definition of fair?
I pose this question to you. If "greed, fascism, xenophobia and armed conflict" is so much worse, why do people choose that? A certain amount of greed leads to successful businesses. Fascism is popular nowadays because democracy has failed the working man. Xenophobia is bred in the bone - curing genetic diseases is tough enough. And as far as armed conflict, well every day I am grateful that the brave soldiers of Ukraine are standing up to Putin's minions so that I don't have to do it in the streets of Boston. And yet, the soldiers they are fighting also believe their cause if just. Again I say, it would help to run some serious simulations that try to answer the question of why people go for these solutions.
My worry is that if you go beyond the platitudes and actually try to put into place the things you espouse, the cure would be as bad as the disease. Think of the people who stuck with diet to cure ulcers or targeting Alzheimer's plaques. Many of them meant well - not all were charlatans.
Before you go espousing things, I suggest you run some simulations.
By the way, I am currently working on a book about meta-ethics. Being a computer techie with a bachelor's in phyiscs, I am doing stuff like modeling the is-ought distinstion as a stress-energy tensor. There is no such thing as a pure virtue or vice. Aristotle was right - there is a Golden Mean.
And having worked in Machine Learning, there is a fundamental theorem in Ensemble Learning (think: society of weak learners) that shows that if your weak learners operate at better than chance (right 51% of the time), then the ensemble will Boost the results of the weak learners. Consider a dictator (Xi Jinping, for example). A dictator is the Ensemble method that combines the efforts of the individuals. If his direction is better than chance, the society functions more efficiently than a herd-of-cats democracy, because democracy is inefficient. On the other hand, if the dictator is worse than chance (Vladimir Putin, for example) then society goes into the ditch very fast. That is why fascism is chosen so often, but democracy wins in the end. If American Democracy is sound, it will be self-correcting after this administration.