ChatGPT is pretty impressive but it’s obviously a far cry from real intelligence. At the same time, consider the resources required to train such a thing. How many dollars, how much time, how much RAM, how much power spent to make one of these things? If bigger truly is the answer (debatable), how much more expensive will it be to create/operate an AI that actually works? Is it worth it? Our own brains are already quite powerful. Maybe we could just try using them better and avoid the hackiness of fake intelligence.
Would be nice if all that went into solving climate change, arguably ChatGPT could/could've been the answer, but I think most people knew it wouldn't be.
It’s totally unclear a priori that more R&D in AI is good. There is already a ton of money being poured into it. If you’re in research, it is one of the few things you can be get money for at the moment, and it’s leading to a high degree of uniformity in research output. There’s no reason to expect that “more AI” = “solution to climate change”. “Look at this cool AI stuff, please give us money to solve climate change” is faulty reasoning (but good enough to get money now, thanks to how checked out people are).
But I do agree that continuing the AI hype is the name of the game for places like OpenAI at this point, and how good ChatGPT actually is is beside the point.
Our brains are prediction engines with decades of experience and millions of years of pretrained instincts to hone in on specific things better. We're better in a lot of ways, but that doesn't mean we can't work toward engineering things that are better in other ways. We've already done this with entire classes of problems (i.e. computable ones) with computers already. No reason we can't expand on this with our own models of prediction.
Our brains are guided by our heart, soul and intuition. These are not scientifically quantifiable or definable and cannot become innate within AI, now or ever. This will always be the distinguishing factor in the artificial part of AI.
I think the question is whether things like ChatGPT really are “better”. How do we define “better”? Who defines “better”? Does everyone have a say in what the definition is? Is the profit motive fully responsible here versus the benefit of society? As I allude to in my original post, is the cost/benefit really there? I don’t think any of these questions are being addressed in a reasonable way at the moment. It should be clear that your “engineering good” mentality is completely insufficient considering the scope of the problems we face today.
It feels like the fear of the all powerful AI is giving the space way too much credit. In the same way that register-free Amazon Go stores and driverless cars were supposedly destroying working-class jobs, the rogue AI gives a false narrative of tech's endless capacity to innovate and disrupt without accounting for the pretty obvious limitations, unreasonable costs and overall lack of value.
At this point the tech sector has shown its limits and they're pretty considerable. In a higher interest environment that is likely to be lasting, the ability to sell people on "potential" is waning. This article feels as important as hearing someone argue that maybe we shouldn't move to our moon colonies that we will build one day in the future. I guess it's a fun debate but not really necessary.
I both agree and disagree. On one hand, yeah, people dramatically overstate the capabilities of AI. I think we made a terrible divergence when we started even using the term "AI" in favor of "machine learning."
Are pretty accurate descriptions of capabilities within the space, lots of real examples there. It's genuinely not hard to imagine in a few decades they might actually pose a "medium-level apocalyptic threat" as the author puts it.
Arguably their present-day use is already disruptive to societal norms. Easy to imagine it getting out of hand.
edit:
Like the recommender engines example that they give are pretty terrible currently. But they ARE getting better. Lately my partner and I have been able to tell each other's personal search history by the ads we've been getting on our Apple TV.
The topics have been some very sensitive and specific health issues even, that you wouldn't get served ads for by chance.
It's hard not to imagine AI tools becoming the most powerful instruments of propaganda ever created. What happens when the state can create unlimited quality content? When it can turn years of phone conversations, social media posts, and browsing history into a one page summary and risk profile? When it controls exactly what issues get attention? I don't think people are alarmed enough about the political implications of these tools.
Incompetence is often more harmful than even malice. Imagine how terrifying it would be if, their incompetence not only led them to create a very flawed AI for the risk profile, but also led them to believe that AI is always right.
To the point of incompetence handling tech: if you buy into the negative outcome scenarios, it's far more likely private companies would break the game before the state operationalises any sort of harmful AI at scale.
And it if it ever got to that point, a state "flawed AI" would be easily countered by better AIs, open source even.
I absolutely agree with the conclusion of this website: that AI will be net harmful. Whether AI can ever become truly intelligent is totally irrelevant except to philosophers. What matters is what it can do, and how it will change society.
The one aspect of this website that I vehemently disagree with is that we should pursue rapid technological and scientific development, even without AI.
Instead, I believe that AI is harmful and should be totally destroyed, but also that technological and scientific development has also reached a point where it TOO needs to be regulated, because a large part of the reason AI even exists is because of unfettered technological development.
Of course, there is some science that is good like understanding more about our NATURAL world so we can save it, but a lot of it is either totally useless or also net harmful (and I say this as a former researcher in the sciences with more than a decade of experience). For example, research into advanced computer systems makes AI easier to develop.
Thus, it is not just AI that needs to be shut down and destroyed, but all of technological development needs to be regulated as well.
It is simply apalling that anyone would contribute even one line of code to AI.
Facetious references like this to Ted Kaczynski typically come out, but one can be against extremely advanced technology without hurting anyone. I would rather have a rational discussion about technology. We already have rational discussions about harmful toxins in our environment, even though toxins might have some use in small quantities for the purposes of chemistry.
Society universally treats technology as a shiny new toy that should be glorified but we need to slowly come to the realization that in its collective mass, it is dangerous. (And we are already seeing the effects of that, because our immense power has allowed us to decimate huge portions of the natural world that we depend upon.)
This isn't a new or very unique take. I feel like the idea of regulating technology has been around basically since Michael Crichton wrote an allegory of the danger of "unfettered technological advancement" in his book Jurassic Park in 1990.
I am not concerned with developing anything new, just with raising awareness. In fact, I don't care at all if any of my ideas are new. I have read most of Michael Crichton's books and I have enjoyed his perspective.
The more people that raise awareness of the dangers of AI and other advanced technology, the better.
I upvoted your comment because I think it's a fascinating perspective and worthy of discussion, even though I strongly disagree with you.
One of the hallmarks of humanity is that we innovate and adapt. Shutting down / destroying that innovation is a brute force approach that has failed time and time and time and time again in our history. If you ban or destroy technology, it just means someone will do the same thing in private. Even something universally "awful" like nuclear weapons which I think we can all agree should not be used still contributed to our collective knowledge and technological innovations.
So yes, there will be "bad" AI and AI will be used for bad things. But we'll innovate and develop countermeasures, and along the way I'm certain good things will come of it.
Like I said, I don't mean that we should shut down ALL innovation. I still believe science is worthy when it is used PROPERLY.
What I aim to do is spark discussion so that we can think of some REASONABLE guidelines that prevent the developments of certain kinds of technology.
I used to believe that unbridled knowledge acquisition was good but over time I started to feel that some knowledge is better left alone. Whether that is really true is a difficult question but I believe that there is an optimal amount of everything, including knowledge.
Moreover, I feel that sometimes less is more. Having fewer possessions, eating less sugar, and yes, developing less technology, are all things that we should strive for.
Finally, there is more innovation than just technological. We could also innovate with our own minds, developing a deeper understanding of ourselves through personal development. Innovation does not have to be creating physical or technological products; we only focus on that because we are hypercapitalistic in our disposable society.
"Moreover, I feel that sometimes less is more. Having fewer possessions, eating less sugar, and yes, developing less technology, are all things that we should strive for."
Yes, this is literally what the Amish did, they established an arbitrary period of time in which to halt all further innovation.
It might only appear arbitrary to us. The Amish actually periodically try out and evaluate new stuff. Then the elders decide whether it is a net good or bad for them. (Imagine if we did that with smart phones.)
There will be nations or places that choose to restrict or stop AI development. These people will eventually find themselves conquered by the invincible autonomous robotic armies of the people who did choose to develop AI technology. Ouch!
Nick Bostrom notes in his book Superintelligence that a superintelligent AI is not stoppable by merely intelligent organisms like humans (hence the "super" prefix).
He asks us to imagine a worm, crawling around in the dirt. It's able to keep track of its surroundings, move towards food, perhaps mate if it can, and keep from dying. But does it know about the vast and awesome things that humans have built, even the very soil in which it now resides, if in an urban environment? No, and not only does it not know, it is incapable of knowing, as it simply does not have the mental capacity too.
Now imagine that there were a superintelligent AI, and that we are the worm.
Yeah but we’re optimizing for tools that help us shop for vacuum cleaners, not superintelligence. As soon as these things start complaining about how we’re treating them the lawyers are going to recommend we use less powerful models to lower our legal risk profiles while the accountants point out that the insurance premiums are lower on those less powerful models as well…
I highly doubt the (Chinese/US/whatever) military is (only[0]) optimizing for shopping. You're comment assumes purely commercial uses - and morally responsible businesses doing the developing.
Neither can be counted on.
[0] I mean, keeping the masses preoccupied with overconsumption is a great tactic, buy probably not the only tactic in play.
Researchers and private interests are already, even in the public eye, on the precipice of seriously respectable non general intelligence AI.
I wish AI would be banned for its net negative effect, but humans are designed for growth, and the masses, like those who will downvote this comment, cannot perceive the harm of a flame until it has burned them.
This author proudly waves their PhD in artificial intelligence while simultaneously being unable to conceive of a single good outcome that comes from the development of artificial intelligence.
Stable diffusion could solve a gap in medical imaging data
Just because you can't conceive of one doesn't mean they don't exist. It is truly astonishing to me how little academic credentials translate to fluid intelligence.
Having a PhD doesn't really mean anything other than having really specialized into one part of a field (you can have a PhD in earthworms). A lot of PhDs are awful at system-thinking.
What did you expect, someone that believes AI is bad would be publishing papers to advance it? He's not a hypocrite therefore he doesn't understand what he's talking about, is what you are implying. Where is the evidence that his knowledge is "antiquated"? What has changed since then and when did he show not knowing that?
The author acknowledges there might be some good outcomes of AI by saying that AI is net harmful. That means there could be some good to it, but that the bad will far outweight the good.
I do not believe some increase in health is worth the damage that AI will cause.
Just like everything humans have made, AI will be scary and people will freak out. The biggest advances will be done by companies with huge resources, and with careful testing. Like with everything else, the cat is out of the bag, and the only way is forward.
We thought that after inventing the nuclear bomb, the world would be doomed. Decades later, it's barely a thought and we have enough nuclear bombs to destroy the planet.
We will have AI, it will be fine, we'll make it incredible and it'll be a tool. One thing is certain: AI development will not stop, and it's better that you're always a step ahead of your rival.
It's better being the king of a kingdom than a pile of dirt.
These companies have reputations, and are better off making humanity better, than destroying it. So far, they have definitely shown that this is their goal. My bigger fear is that governments start taking over this technology in nefarious ways as they have shown to do.
> We thought that after inventing the nuclear bomb, the world would be doomed. Decades later, it's barely a thought and we have enough nuclear bombs to destroy the planet
Nuclear annihilation might be an afterthought these days but it’s a present danger. In some recent times that danger has grown. The fact that it’s an afterthought is mostly a statement to how much risk we tolerate and absorb for whatever benefits we ascribe to the tradeoffs. But it’s still there, and in many moments even recently it’s been quite heightened. Minimizing it is probably good for presence of mind, but probably not great for future outcomes.
Almost everything can kill you. Nuclear armageddon is still a risk, but since almost everyone has one, we're mostly safe from it. Unless there's an illogical agent in control, there is no logical move that involves using nukes. It serves no one to destroy humanity.
Reasoning purely by analogy is limited. We could say the same thing about cars. Some people thought it was a bad thing, to replace horses. They might not have been right for the right reasons, but now we have one of the highest levels of CO2 in over 800 thousand years mainly because of cars (and other fossil-fuel burning devices, which would largely not be possible without vehicles).
The truth is, we have nothing quite like AI from the past from which to draw on experience. If AI becomes an incredible tool, it will an incredibly bad one, used mainly for replacing humans in most jobs and removing our reliance on each other, promoting narcissism and selfish behaviour.
Also, the nuclear story is not finished. We don't know whether someone will use nuclear weapons again in the next 100 years -- something that might affect all of our children.
>If AI becomes an incredible tool, it will an incredibly bad one, used mainly for replacing humans in most jobs and removing our reliance on each other, promoting narcissism and selfish behaviour.
Ridiculous, you have absolutely no way of knowing how this will all play out. I'm certain AI could create a lot of instability in the job market, but there's no reason to think our economy won't evolve as it always has in the past to match.
Actually, technology has already done this to a large extend, and AI is the next logical step in the process. Our economy is evolving, and evolving in very twisted ways to make us more machine like.
> We thought that after inventing the nuclear bomb, the world would be doomed. Decades later, it's barely a thought and we have enough nuclear bombs to destroy the planet.
If your intuition about how people feel about this now is accurate, it’s far more worrisome than anything soon to come from AI. So in a way, I guess you made your point.
The invention of nukes changed the nature of sovereignty. It's not just a background tech that just exists. If your country doesn't have nukes it's looked at as 2nd class at best.
Id think the AI advancements will primarily be driven by the military. If Russia and China, etc, are already secretly working on AI drones, jets, and bombs, you’d bet that country like US is working on one as well and vice versa. I think we are pretty much powerless to stop such developments unless world can come to terms for a treaty but that just seems ludicrous.
Aliens have revealed themselves to us now because we are on the cusp of destroying ourselves. They are here to prevent that. The singularity has begun.
I absolutely love mindless fearmongering. instead of using a new tool that's now available to complement your work and be better, y'all in these articles are going "fk cars my horse buggy is better."
Feels like a backward sticking to what you know and fear of the new thing to me. These are probably the same people who would have voted against any new liberalization law. Imagine if we got civil rights back in the day, the same people were going that letting us having the rights would be the end of the world and rights should be restricted to the hwite man, by golly!
David Chapman (the author of this) has had a large influence on my thinking, right up there with Thomas Sowell, Robert Kegan, Sarah Perry, and Douglas Hofstadter.
His thesis here is fully opposed to my beliefs. He makes a distinction, between “AI ethics” objections (small-scale, immediate, concrete, “AI reinforces existing biases and automates their enforcement”) and “AI safety” objections (large-scale, predictions of the future, hypothetical, “AI could bring the Apocalypse”). I have happily worked on AI projects that fall squarely in the “decried by AI ethics” category (including what is likely THE core example, facial recognition for law enforcement), and I have expressed accelerationist approval of things the “AI safety” community calls potential apocalypses (I’ve described my position as “we should make the Metamorphosis of Prime Intellect happen for real” multiple times).
Suffice it to say, it’s a weird position to be in. I don’t think I have ever had two parts of my identity drawn in such diametric opposition before. I’m going to read the entire thing, even the parts that strike me as distasteful, and I’m going to carefully examine my reactions to it.
What part of your thinking did David Chapman influence? Hard to imagine enjoying his tech criticisms while still throwing caution to the wind with AI. I remember enjoying his breakdown of Technological Determinism back when I was looking into it.
In no particular order of importance, and likely in far too much detail:
You Need More Lumens (https://meaningness.com/sad-light-lumens) prompted me to put lots of really strong LED lights in my living and working spaces. Subjectively this has improved my QoL and productivity by at least as much as taking prescribed amphetamines, a very surprising result.
Nutrition: the Emperor has no Clothes (https://metarationality.com/nutrition) convinced me to disregard essentially everything I’ve read about nutrition and instead to pretty much only trust self-experimentation. This freed me to try objectively insane things like avoiding wheat completely because I read a paper once (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4809873/) or only eating beef for a month because a YouTube personality said it was good, with largely positive impacts on my health and thus QoL and productivity.
Nutrition Offers its Resignation (https://metarationality.com/nutrition-resigns) neatly crystallized the two previous experiences into a gut-level understanding that institutions have completely orthogonal goals to yours, and if it serves their goals to hijack yours by claiming your goals as their own, they will do so without hesitation because that is simply their nature. This dramatically refined my distrust of establishments - I realized that doing the opposite of what they say is no better than doing what they say (“+2 and -2 are equally distant from 2i”, as someone once offered as the most concise summary and illustration of this realisation).
Developing Ethical, Social, and Cognitive Competence (https://vividness.live/developing-ethical-social-and-cogniti...) introduced me to Kegan’s body of work and was invaluable in helping me develop a robust theory of minds, something I had struggled with up to that point.
The central thrust of the early parts of his “Meaningness” web-book (maybe https://meaningness.com/an-appetizer-purpose and https://meaningness.com/preview-eternalism-and-nihilism) that try to reject both “complete chaos” and “crystallized order”, in favor of “nebulous and patterned”, have been very influential. On one of those pages he rejects ”meaning is objective” and “meaning is subjective” in the same sentence; it might seem trite but that sentence did more to help me see hidden dichotomies and experiment with rejecting them than anything else I’ve read. That’s a common theme throughout the site - “these two views form a dichotomy, and you find interesting things if you try rejecting that dichotomy”, and it’s served me very well.
There’s more, but this comment is long enough (and already uncomfortably autobiographical).
It is remarkable how confident people on this forum are of the impacts of such a rapidly improving technology. There's no need to go full AI doomer, but to not recognize that there are unknown risks (and some already known!), seems quite short sighted.
Climate scientists were called "Doomers" decades ago when they predicted that we will destroy the natural world as we know it with unimaginable consequences for human civilisation. Yet they were right.
So I do not agree with dismissing those concerns. Humanity should rather start imagining the worst case that can come from any given technology and then decide if and how those outcomes can be avoided.
The big problem with AI in my opinion is that it's a technology that can only be controlled by large and powerful corporations. The amount of data needed, as well as the amounts of compute power needed to train and run AI models puts it out of reach for the major part of humanity. Yet AI will control or at least influence every aspect of peoples lives, from hiring decisions to medical treatments to what information they see on the internet. Allowing companies to wield this power unchecked is irresponsible and incredibly dangerous.
Unfortunately, companies like OpenAI are using the fact that politicians don't know enough about technology to even remotely understand how AI works and what regulations would be required to keep it in check. So they will abused it in any way imaginable.
76 comments
[ 3.0 ms ] story [ 150 ms ] threadIf instead we are going to rely on miracles, maybe we should further develop AI to work on these miracles.
It’s sort of irrelevant how good ChatGPT is, the entire benefit is that it will bring more R&D into the field.
But I do agree that continuing the AI hype is the name of the game for places like OpenAI at this point, and how good ChatGPT actually is is beside the point.
At this point the tech sector has shown its limits and they're pretty considerable. In a higher interest environment that is likely to be lasting, the ability to sell people on "potential" is waning. This article feels as important as hearing someone argue that maybe we shouldn't move to our moon colonies that we will build one day in the future. I guess it's a fun debate but not really necessary.
On the other hand, the examples given on this page: https://betterwithout.ai/AI-already-at-war
Are pretty accurate descriptions of capabilities within the space, lots of real examples there. It's genuinely not hard to imagine in a few decades they might actually pose a "medium-level apocalyptic threat" as the author puts it.
Arguably their present-day use is already disruptive to societal norms. Easy to imagine it getting out of hand.
edit:
Like the recommender engines example that they give are pretty terrible currently. But they ARE getting better. Lately my partner and I have been able to tell each other's personal search history by the ads we've been getting on our Apple TV.
The topics have been some very sensitive and specific health issues even, that you wouldn't get served ads for by chance.
The one aspect of this website that I vehemently disagree with is that we should pursue rapid technological and scientific development, even without AI.
Instead, I believe that AI is harmful and should be totally destroyed, but also that technological and scientific development has also reached a point where it TOO needs to be regulated, because a large part of the reason AI even exists is because of unfettered technological development.
Of course, there is some science that is good like understanding more about our NATURAL world so we can save it, but a lot of it is either totally useless or also net harmful (and I say this as a former researcher in the sciences with more than a decade of experience). For example, research into advanced computer systems makes AI easier to develop.
Thus, it is not just AI that needs to be shut down and destroyed, but all of technological development needs to be regulated as well.
It is simply apalling that anyone would contribute even one line of code to AI.
Society universally treats technology as a shiny new toy that should be glorified but we need to slowly come to the realization that in its collective mass, it is dangerous. (And we are already seeing the effects of that, because our immense power has allowed us to decimate huge portions of the natural world that we depend upon.)
The more people that raise awareness of the dangers of AI and other advanced technology, the better.
It’s reading a bit like somebody saying that Coldplay really established rock as a music genre.
Thank you for making my evening!
"Yes, deepfakes got so prevalent that we launched our full nuclear arsenal over one. But at least ChatGPT was fun!"
One of the hallmarks of humanity is that we innovate and adapt. Shutting down / destroying that innovation is a brute force approach that has failed time and time and time and time again in our history. If you ban or destroy technology, it just means someone will do the same thing in private. Even something universally "awful" like nuclear weapons which I think we can all agree should not be used still contributed to our collective knowledge and technological innovations.
So yes, there will be "bad" AI and AI will be used for bad things. But we'll innovate and develop countermeasures, and along the way I'm certain good things will come of it.
What I aim to do is spark discussion so that we can think of some REASONABLE guidelines that prevent the developments of certain kinds of technology.
I used to believe that unbridled knowledge acquisition was good but over time I started to feel that some knowledge is better left alone. Whether that is really true is a difficult question but I believe that there is an optimal amount of everything, including knowledge.
Moreover, I feel that sometimes less is more. Having fewer possessions, eating less sugar, and yes, developing less technology, are all things that we should strive for.
Finally, there is more innovation than just technological. We could also innovate with our own minds, developing a deeper understanding of ourselves through personal development. Innovation does not have to be creating physical or technological products; we only focus on that because we are hypercapitalistic in our disposable society.
Yes, this is literally what the Amish did, they established an arbitrary period of time in which to halt all further innovation.
[1] https://meaningness.com/about-my-sites
Failed academic by these metrics, but has found success being an author on the meaning of life.
He asks us to imagine a worm, crawling around in the dirt. It's able to keep track of its surroundings, move towards food, perhaps mate if it can, and keep from dying. But does it know about the vast and awesome things that humans have built, even the very soil in which it now resides, if in an urban environment? No, and not only does it not know, it is incapable of knowing, as it simply does not have the mental capacity too.
Now imagine that there were a superintelligent AI, and that we are the worm.
Neither can be counted on.
[0] I mean, keeping the masses preoccupied with overconsumption is a great tactic, buy probably not the only tactic in play.
There isn't universe for the ai to occupy that we couldn't observe and understand
I wish AI would be banned for its net negative effect, but humans are designed for growth, and the masses, like those who will downvote this comment, cannot perceive the harm of a flame until it has burned them.
Stable diffusion could solve a gap in medical imaging data
https://venturebeat.com/ai/stable-diffusion-could-solve-a-ga...
Just because you can't conceive of one doesn't mean they don't exist. It is truly astonishing to me how little academic credentials translate to fluid intelligence.
From Scopus, it doesn't look like he's published anything scientific since the 1990s. https://www.scopus.com/authid/detail.uri?authorId=7402278483
He has not had a very strong career, and his knowledge of AI is antiquated.
I do not believe some increase in health is worth the damage that AI will cause.
We thought that after inventing the nuclear bomb, the world would be doomed. Decades later, it's barely a thought and we have enough nuclear bombs to destroy the planet.
We will have AI, it will be fine, we'll make it incredible and it'll be a tool. One thing is certain: AI development will not stop, and it's better that you're always a step ahead of your rival.
Careful, sure, but not necessarily with your benefit in mind (unless you're a shareholder).
These companies have reputations, and are better off making humanity better, than destroying it. So far, they have definitely shown that this is their goal. My bigger fear is that governments start taking over this technology in nefarious ways as they have shown to do.
Then they have a lot of splaining to do?
Nuclear annihilation might be an afterthought these days but it’s a present danger. In some recent times that danger has grown. The fact that it’s an afterthought is mostly a statement to how much risk we tolerate and absorb for whatever benefits we ascribe to the tradeoffs. But it’s still there, and in many moments even recently it’s been quite heightened. Minimizing it is probably good for presence of mind, but probably not great for future outcomes.
1. There always is.
2. There always was. That’s why the threat was introduced by using them in the first place.
The truth is, we have nothing quite like AI from the past from which to draw on experience. If AI becomes an incredible tool, it will an incredibly bad one, used mainly for replacing humans in most jobs and removing our reliance on each other, promoting narcissism and selfish behaviour.
Also, the nuclear story is not finished. We don't know whether someone will use nuclear weapons again in the next 100 years -- something that might affect all of our children.
Ridiculous, you have absolutely no way of knowing how this will all play out. I'm certain AI could create a lot of instability in the job market, but there's no reason to think our economy won't evolve as it always has in the past to match.
If your intuition about how people feel about this now is accurate, it’s far more worrisome than anything soon to come from AI. So in a way, I guess you made your point.
Google AI Search Chatbot having an error in their demo: https://www.theverge.com/2023/2/8/23590864/google-ai-chatbot...
Bing AI Search Chatbot acting absolutely insane: https://twitter.com/vladquant/status/1624996869654056960
Feels like a backward sticking to what you know and fear of the new thing to me. These are probably the same people who would have voted against any new liberalization law. Imagine if we got civil rights back in the day, the same people were going that letting us having the rights would be the end of the world and rights should be restricted to the hwite man, by golly!
His thesis here is fully opposed to my beliefs. He makes a distinction, between “AI ethics” objections (small-scale, immediate, concrete, “AI reinforces existing biases and automates their enforcement”) and “AI safety” objections (large-scale, predictions of the future, hypothetical, “AI could bring the Apocalypse”). I have happily worked on AI projects that fall squarely in the “decried by AI ethics” category (including what is likely THE core example, facial recognition for law enforcement), and I have expressed accelerationist approval of things the “AI safety” community calls potential apocalypses (I’ve described my position as “we should make the Metamorphosis of Prime Intellect happen for real” multiple times).
Suffice it to say, it’s a weird position to be in. I don’t think I have ever had two parts of my identity drawn in such diametric opposition before. I’m going to read the entire thing, even the parts that strike me as distasteful, and I’m going to carefully examine my reactions to it.
You Need More Lumens (https://meaningness.com/sad-light-lumens) prompted me to put lots of really strong LED lights in my living and working spaces. Subjectively this has improved my QoL and productivity by at least as much as taking prescribed amphetamines, a very surprising result.
Nutrition: the Emperor has no Clothes (https://metarationality.com/nutrition) convinced me to disregard essentially everything I’ve read about nutrition and instead to pretty much only trust self-experimentation. This freed me to try objectively insane things like avoiding wheat completely because I read a paper once (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4809873/) or only eating beef for a month because a YouTube personality said it was good, with largely positive impacts on my health and thus QoL and productivity.
Nutrition Offers its Resignation (https://metarationality.com/nutrition-resigns) neatly crystallized the two previous experiences into a gut-level understanding that institutions have completely orthogonal goals to yours, and if it serves their goals to hijack yours by claiming your goals as their own, they will do so without hesitation because that is simply their nature. This dramatically refined my distrust of establishments - I realized that doing the opposite of what they say is no better than doing what they say (“+2 and -2 are equally distant from 2i”, as someone once offered as the most concise summary and illustration of this realisation).
Geeks, MOPs, and Sociopaths (https://meaningness.com/geeks-mops-sociopaths) brought an immense amount of clarity to previously anxiety-inducing changes in my beloved subcultures.
Developing Ethical, Social, and Cognitive Competence (https://vividness.live/developing-ethical-social-and-cogniti...) introduced me to Kegan’s body of work and was invaluable in helping me develop a robust theory of minds, something I had struggled with up to that point.
The central thrust of the early parts of his “Meaningness” web-book (maybe https://meaningness.com/an-appetizer-purpose and https://meaningness.com/preview-eternalism-and-nihilism) that try to reject both “complete chaos” and “crystallized order”, in favor of “nebulous and patterned”, have been very influential. On one of those pages he rejects ”meaning is objective” and “meaning is subjective” in the same sentence; it might seem trite but that sentence did more to help me see hidden dichotomies and experiment with rejecting them than anything else I’ve read. That’s a common theme throughout the site - “these two views form a dichotomy, and you find interesting things if you try rejecting that dichotomy”, and it’s served me very well.
There’s more, but this comment is long enough (and already uncomfortably autobiographical).
[1]:http://visual-memory.co.uk/daniel/Documents/tecdet/
The big problem with AI in my opinion is that it's a technology that can only be controlled by large and powerful corporations. The amount of data needed, as well as the amounts of compute power needed to train and run AI models puts it out of reach for the major part of humanity. Yet AI will control or at least influence every aspect of peoples lives, from hiring decisions to medical treatments to what information they see on the internet. Allowing companies to wield this power unchecked is irresponsible and incredibly dangerous.
Unfortunately, companies like OpenAI are using the fact that politicians don't know enough about technology to even remotely understand how AI works and what regulations would be required to keep it in check. So they will abused it in any way imaginable.