If performance indeed asymptotes, and if we are not at the end of silicon scaling or decreasing cost of compute, then it will eventually be possible to run the very best models at home on reasonably priced hardware.
Eventually the curves cross. Eventually the computer you can get for, say, $2000, becomes able to run the best models in existence.
The only way this doesn’t happen is if models do not asymptote or if computers stop getting cheaper per unit compute and storage.
This wouldn’t mean everyone would actually do this. Only sophisticated or privacy conscious people would. But what it would mean is that AI is cheap and commodity and there is no moat in just making or running models or in owning the best infrastructure for them.
Tangent but... I kinda like the Python language. What I don't like about Python is the way environments are managed.
This is something I generally believe, but I think it's particularly important for things like languages and runtimes: the idea of installing things "on" the OS or the system needs to die.
Per-workspace or per-package environment the way Go, Rust, etc. does it is correct. Installing packages globally is wrong.
There should not be such a thing as "globally." Ideally the global OS should be immutable or nearly so, with the only exception being maybe hardware driver stuff.
(Yes I know there's stuff like conda, but that's yet another thing to fix a fundamentally broken paradigm.)
> This is something I generally believe, but I think it's particularly important for things like languages and runtimes: the idea of installing things "on" the OS or the system needs to die.
Python has been trying to kill it for years; or at least, the Linux distros have been seeking Python's help in killing it on Linux for years. https://peps.python.org/pep-0668/ is the latest piece of this.
I feel like this principle could be codified as "the system is not a workspace."
The use of the system as a workspace goes back to when computers were either very small and always personal only to one user, or when they were very big and administrated by dedicated system administrators who were the only ones with permission to install things. Both these conditions are obsolete.
But the system is not a workspace acts like resources are free. Everything that’s wrong with a modern computer being slower than one from 30 years ago at running user applications has its roots in this kind of thing. It’s more obvious on mobile devices but desktops still suffer. Android needs more RAM and had worse power utilization until a lot was done to move toward native compiled code and background process control. Meanwhile Electron apps think it’s okay to run multiple copies of Javascript environments like working RAM is free and performance isn’t hurt.
Totally agreed. I read somewhere that the only place these features help is sports. They should not be defaults. They make shows and films look like total crap.
Actually, they do not belong anywhere. If you look at the processing pipeline necessary to, for example, shoot and produce modern sporting events in both standard and high dynamic range, the last thing you want is a television that makes its own decisions based on some random setting that a clueless engineer at the manufacturer thought would be cool to have. Companies spend millions of dollars (hundreds of millions in the case of broadcasters) to deliver technically accurate data to televisions.
These settings are the television equivalent of clickbait. They are there to get people to say "Oh, wow!" at the store and buy it. And, just like clickbait, once they have what they clicked on, the experience ranges from lackluster and distorted to being scammed.
As someone who has built multi-camera live broadcast systems and operated them you are 100% correct. There is color correction, image processing, and all the related bits. Each of these units costs many times more and is far more capable with much higher quality (in the right hands) than what is included in even the most high end TV.
I speak from experience. I spend approximately twenty years developing technology for broadcast, motion picture, production and post-production. That also included systems integration, where we designed and built all kinds of facilities. The largest I was personally involved with had a $65M budget.
Most people have absolutely no idea what goes into making the pixels on their screens flicker with quality content.
They're the equivalent of the pointless DSP audio modes on 90's A/V receivers. Who was ever going to use "Concert Hall", "Jazz Club", or "Rock Concert" with distracting reverb and echo added to ruin the sound.
I think it is helpful to have settings that you can change, although the default settings should probably match those intended by whoever made the movie or TV show that you are watching, according to the specification of the video format. (The same applies to audio, etc.)
This way, you should not need to change them unless you want nonstandard settings for whatever reason.
In previous eras there were many purists who considered photography not-art, sequencer and synthesizer made music not-music, other forms of (non-AI) digital art less legitimate than their more manual classical counterparts, etc. This is the same discourse all over again.
Is electronic music where the artist composes it on a screen and then hits 'play' music? I think it is, of course, but I have had experiences where I went to see a musician "live" and well... they brought the laptop with them. But I think it still counts. It was still fun.
AI slop is to AI art what point and shoot amateur photography is to artistic photography. The difference is how much artistic intent and actual work is present. AI art has yet to get people like Ansel Adams, but it will -- actual artists who use AI as a tool to make novel forms and styles of art.
Anti-photography discourse sounds exactly like anti-AI discourse to the point that you could search and replace terms and have the same rants.
Another thing I expect to see is novelists using AI to create at least passable live action versions of their stories. I don't think these will put real actors or actresses out of work for a long time, but I could see them serving as "sizzle reels" to sell a real production. If an author posts their AI-generated film of their novel and it gets popular, I could see a studio picking it up and making a real movie or TV show from it.
> Is electronic music where the artist composes it on a screen and then hits 'play' music?
If X composes something, X is an artist. The person playing a composed work is a performer. Some people have both the roles of artist and performer for a given work.
To say an AI composes something is anthropomorphizing a computer. If you enter a prompt to make a machine generate work based on existing artists' art, you're not composing (in the artistic sense) and neither is the computer. Math isn't art even if it's pretty or if mathematical concepts are used in art.
The term "director" instead of composer or artist conveys what's happening a lot better with telling machines to generate art via prompts.
I mostly agree with your sentiment, but saying "math is not art" is the same as saying "writing is not art". Calculation isn't art. But math isn't calculation. Math is a social activity shared between humans. Like writing, much of it is purely utilitarian. But there's always an aesthetic component, and some works explore that without regard to utility. It's a funny kind of art, accessible to few and beautiful to even fewer. But there is an art there.
Incorrect. Art is practice. It's literally what the word means historically. Put in "Etymology of the word 'art'" in your favorite search engine or LLM.
If someone is entering a prompt to generate an image in a model I have access to, I don't really need to pay them to do it, and definitely don't need to pay them as much to do it as I would an actual artist, so it is deceptive for them to represent themselves as someone who could actually draw or paint that. If the product is what counts then truth in advertising is required so the market can work.
The vast majority of artists in all fields don't really have their own style and are just copying other people's. Doesn't matter whether we're talking about art, literature, music, film, whatever.
It takes a rare genius to make a new style, and they come along a few times a generation. And even they will often admit they built on top of existing styles and other artists.
I'm not a fan of AI work or anything, but we need to be honest about what human 'creativity' usually is, which for most artists is basically copying the trends of the time with at most a minor twist.
OTOH, I think when you start entering the fringes of AI work you really start seeing how much it's just stealing other people's work though. With more niche subjects, it will often produce copies of the few artists in that field with a few minor, often bad, changes.
Sure, you can say that AI is just "stealing like an artist", but that makes the AI the artist in this scenario, not the prompter.
It bothers me that all of the AI "artists" insist that they are just the same as any other artist, even though it was the AI that did all of the work. Even when a human artist is just copying the styles they've seen from other artists, they still had to put in the effort to develop their craft to make the art in the first place.
I'm not against AI art per se, but at least so far, most “AI artists” I see online seem to care very little about the artistry of what they’re doing, and much much more about selling their stuff.
Among the traditional artists I follow, maybe 1 out of 10 posts is directly about selling something. With AI artists, it’s more like 9 out of 10.
It might take a while for all the grifters to realize that making a living from creative work is very hard before more genuinely interesting AI art starts to surface eventually. I started following a few because I liked an image that showed up in my feed, but quickly unfollowed after being hit with a daily barrage of NFT promotions.
I don't believe that there is near enough room for creativity to shine through in the prompt-generation pipeline, and I find the mention of a talent like Ansel Adams in this context asinine. There is no control there, and without control over creation I don't believe that creativity CAN flourish, but I may be wrong.
Electronic music is analogous to digital art made by humans, not generated art.
Defining art in this way is like defining intelligence as the possession of a degree from Stanford. It's just branding.
Art shouldn't make you feel comfortable and safe. It should provoke you and in this sense AI art is doing the job better than traditional art at the moment here.
Other than the technological aspect, there's nothing new under the sun here. And at its very worst, AI art is just Andy Warhol at hyperscale.
I think it's actually quite apt to look at all of "AI art" as a single piece, or suite, with a unified argument or theme. Maybe in that sense it is some kind of art, even if it wasn't intended that way by its creators.
Similarly, I'm not sure that argument is making the point those who deploy it intend to make.
I think the entire fear of AI schtick to farm engagement is little more than performance art for our FAANNG overlords personally. It behaves precisely like the right wing manosphere but with different daily talking points repeated ad nauseum. Bernie Sanders has smelled the opportunity here and really stepped up his game.
But TBF, performance art theatre is art as well.
The end game IMO will be incorporation of AI art toolsets into commercial art workflows and a higher value placed on 100% human art (however that ends up being defined) and then we'll find something new and equally idiotic to trigger us or else we might run out of excuses and/or scapegoats for our malaise.
> incorporation of AI art toolsets into commercial art workflows and a higher value placed on 100% human art
I don't even really believe serious artists need to totally exclude themselves from using genAI as a tool, and I've heard the same from real working artists (generally those who have established careers doing it). Unfortunately, that point inhabits the boring ideological center and is drowned out by the screaming from both extremes.
They aren't, but some are already using pseudonyms to experiment with it to avoid the haters condemning them for doing so. And their work is predictably far superior from the get-go to asking Sora to ghiblify your dog.
> Art shouldn't make you feel comfortable and safe. It should provoke you and in this sense AI art is doing the job better than traditional art at the moment here.
jumpscares and weapons being used at others aren't art
I Ghiblified a photo of my dog when chatgpt 4 came out. I was utterly horrified by the results.
It's exciting being able to say that I am an artist, I always wondered what my life would have been had I gone into the arts, and now I can experience it! Thank you techmology.
If you really want to experience the struggles and persecution of an artist, you should empty your bank account and find a life partner to support you while you struggle with your angst and inner trauma that are the source of your creativity. But, to be fair, complaining about AI art is a great start down that path!
How else would you address the incessant ramblings of people who figuratively curse the sunset daily? After AI art has been integrated into the already existing suite of digital art applications (which themselves were once not considered art), whatever shall you complain about next?
Now if you wanted to define art to require 100% bodily fluids and solids 100% handcrafted to be the only real art, now that I'd understand.
You may check these videos by Oleg Kuvaev.
100% generated using AI.
Everything: text, music, characters, voices, editing -- all done via prompts, using multiple engines (I think he mentioned about a dozen services involved).
I would not call it "high art", but it's definitely not a slop, it's an artist skillfully using AI as a tool.
While we're sharing AI generated videos, IGORRR's ADHD music video [0] is definitively art, zero question about it. I don't think typing a prompt in and taking the output as it comes is art -- good art, anyway (the point-and-shoot photography comparison is apt) -- but that doesn't mean AI can't be used to make truly new, creative and unique art too.
This is absolutely slop. Higher quality slop, but slop nonetheless. Ask yourself: what does it say? What does it change in you? How this makes you feel?
Artists use their medium to communicate. More often than not, everything in a piece is deliberate. What is being communicated here? Who deliberated on the details?
Those videos are as much "art" as Marvel's endless slop is "art".
> I don't believe that there is near enough room for creativity to shine through in the prompt-generation pipeline
I mean you are building a prompt and tweaking it. I mean even if you didn't do that you could still argue that finding it is in itself a creative akin to found art [1].
I suppose. You're "finding" something that didn't exist and that nobody ever cared about. Something that you wrote, mashed against the tensors trained on real artist creations, and out came the thing that you "found".
I'm genuinely amazed at how some people perceive art.
To me art has always been "an interesting idea". Decorative things that take skill to me are crafts. Sure, it's a water color of your garden, but what does it tell us about the human condition? Sure, it's skilled... but it's empty. Give me Jackson Pollock or Picasso. Give me a new way to see the world. Pure skill to me is as impressive as cup-stacking personally.
Not saying you have to agree, but it is a distillation of how some portion of the world sees the world.
You know that you can give a drawing as input for image generation, right? I think there's a lot of creativity possible with AI image generation. Things like ControlNet and the various loras and upscale methods etc all add a lot of choice.
How much room for creativity is there with a camera? Angle, lighting, F-stop, film type, film processing? I have a local image generator app called Draw Things that has many times more options than this.
Early synthesizers weren't that versatile either. Bands like Pink Floyd actually got into electronics and tore them apart and hacked them. Early techno and hip-hop artists did similar things and even figured out how to transform a simple record player into a musical instrument by hopping the needle around and scratching records back and forth with tremendous skill.
Serious AI artists will start tearing apart open models and changing how they work internally. They'll learn the math and how they work just like a serious photographer could tell you all about film emulsions and developing processes and how film reacts to light.
Art's never about what it does. It's about what it can do.
> How much room for creativity is there with a camera? Angle, lighting, F-stop, film type, film processing?
How many subjects exist in the world to be photographed? How many journeys might one take to find them? How many stories might each subject tell with the right treatment?
> Serious AI artists will start tearing apart open models and changing how they work internally. They'll learn the math and how they work just like a serious photographer could tell you all about film emulsions and developing processes and how film reacts to light.
I agree that "AI art" as it exists today is not serious.
"AI art" today is mostly play, which is usually the first thing you get with new artistic tools. People just fool around with them in an un-serious way. There's also some porn. Porn is always early. It was one of the first uses for moving pictures, for example.
"The early adopters of new technologies are usually porn and the military." Forget where I heard that but it's largely true.
I do not think that the things you say will happen, will ever happen.
Also, photography has the added benefit of documenting the world as it is, but through the artist's lens. That added value does not exist when it comes to slop.
I don't believe that there is near enough room for creativity to shine through in the prompt-generation pipeline
You seem so sure that you'll always be able to tell what you're looking at, and whether it's the result of prompting or some unspecified but doubtlessly-noble act of "creativity."
Keyboards have had functions that let them play music at the touch of button for decades.
Decades later we still don't consider anyone using that function a musician.
>actual artists who use AI as a tool to make novel forms and styles of art.
writing a prompt lol
We don't compare Usain Bolt to Lewis Hamilton when talking about fastest runners in the world.
But hey think about how much money you could save on a wedding photographer if you just generate a few images of what the wedding probably looked like!
The (wedding) photographer is likely going to use this AI themselves though. They used Photoshop way back in the day to touch up images. They're going to be doing the same with genAI. Content-aware fill is one of the most useful tools they have.
There is a "demo" button on synthesizers that plays a canned melody, therefore playing canned melodies is all synthesizers can do, therefore nobody that uses a synthesizer is a real musician.
> AI slop is to AI art what point and shoot amateur photography is to artistic photography.
Sorry... It's all slop buddy. The medium is the message, and genAI's message is "I want it cheap and with low effort, and I don't care too much about how it looks"
It is more useful to think about it in terms of what that effort actually entails.
If you haven't ever written a novel, or even a short story, you cannot possibly imagine how much of your own weird self ends up in it, and that is a huge part of what will make it interesting for people to read. You can also express ideas as subtext, through the application of technique and structure. I have never reached this level with any form of visual art but I imagine it's largely the same.
A prompt, or even a series of prompts, simply cannot encode such a rich payload. Another thing artists understand is that ideas are cheap and execution is everything; in practice, everything people are getting out of these AI tools is founded on a cheap idea and built from an averaging of everything the AI was trained on. There is nothing interesting in there, nothing unique, nothing more than superficially personal; just more of the most generic version of what you think you want. And I think a lot of people are finding that that isn't, in fact, what they want.
At the very least, art usually contains effort signifiers. Yes, an artist could potentially employ gingerbread men cut from construction paper in a work, but no, construction paper gingerbread men are typically not in the same league as David.
For fun I decided to try out the find and replace on this comment
> Sorry... It's all slop buddy. The medium is the message, and photography's message is "I want it cheap and with low effort, and I don't care too much about how it looks"
Hmm... it seems like you have failed to actually make an argument here
The logical implication of your view is that if someone or something has a halo, they can shit in your mouth and it's "good." The medium is the message, after all.
This is the same pretentious art bullshit that regular people fucking hate, just repackaged to take advantage of public rage at tech bro billionaires.
Whatever, man, this guy isn't wrong. Look at the example he gave how a camera made it so that anyone could do what only a few could. Novel art is just a candid shot now. It forced art to completely change its values. Much of the same will happen now. The difference is that with the past, we still needed artists to take advantage of them while now, it all can be completely automated. It's disgusting but I'm sure purest thought the same of every innovation.
Not "wasted." Handed to allies of the administration. It's just naked kleptocracy.
I suspect that the "ball room" attachment to the White House will also still be a hole by the end of the administration, but a lot of money will get handed out.
I’m shocked that a 256 core Epyc can’t do millions of requests per second at a minimum. Is it limited by the net connection or is there still this much inefficiency?
It almost certainly can, even old intel systems with dual CPU 16 core systems could do 4 and a half million a second [1]. At a certain point network/kernel bottlenecks become apparent though, rather than being compute limited.
Like anything it really depends on what they are doing, if you wanted to just open and close a connection you might run into bottle necks in other parts of the stack before the CPU tops out but the real point is that yea, a single machine is going to be enough.
I think the most likely bottleneck is gonna be your NIC hating getting a ton of packets. Line rate with huge frames is quite different than line rate with just ICMP packets, for instance (see CME binary glink market data for a similarly stressful experience to the ICMP).
All the non-technical people I know loved it. It's pretty. It's neat. It looks cool. Apple is a consumer products company.
My personal feeling on it is just "meh." My productivity with my laptop hasn't changed. I'm not a huge fan but it's not a deal breaker. I still find it better than Windows 11 for the most part, and Linux has other issues as a daily driver for me.
IMHO Apple needs a "tick" release where they only polish and fix bugs and usability issues with an almost total feature freeze. I've heard they may be doing that.
Looking at popularity of similar design on iOS it would be surprising “non-technical” users like it. People HATE new iOS. Low contrast, not clear layering and focus, things being moved around for unknown reasons.
Also who uses MacOs beaides developers? Majority are creative prosumers in arts/design and they are even more annoyed by messed up designs. What you are left with are lawyers, writers, students? I guess they might like it.
I said that in the next sentence. Than again this is really true only in US. In rest of the world (including europe) Macs are seen as luxury environment.
Macs are not seens as a luxury environment in Sweden at least.
Of the people I know only old folks, gamers and some techies own PCs. A lot of people will however just use whatever wintel laptop their employer provides them with.
I'm not arguing that macs are more common than non-macs, only that "In rest of the world (including europe) Macs are seen as luxury environment." is false.
Chromebooks dominate K12 here so it kinda depends on what you mean by "students". Once people start buying their own computers however my impression is that Macs are quite common of not dominating.
Mac have been certainly picking up in usage worldwide but it's been always pretty openly premium luxury product? Especially in the past. It is even marketed like that. They might be changing their reputation since their entry Macbooks are pretty great price/performance but people simply buy cheaper.
Marketshare of MacOS is like 15% worldwide (curiously declining in US). That's a minor platform.
Also stop with Chromebooks. It might dominate schools in US (often mandatory) and it is popular in specific countries like India. But in majority of the world it's absolutely unknown with global marketshare of like 1%.
And I’m not arguing that Mac’s are the common man’s choice. All I said was that the statement about Macs being seen a luxury environment in every country except the US is plainly false as I know at least a couple of other countries where Mac’s are quite common even among the non-affluent classes.
Very few uni and high school students have MacBooks in East and Southeast Europe and it’s seen as quite a flex there. They’re also impractical for those in engineering schools due to required software that only works on Windows.
No, our nature is to satiate our dopamine system. That system evolved to keep us fed, nourished, and to make us make friends and belong and have sex to make more humans. The problem is that we are now so smart and clever that we can start learning how the dopamine system works and hacking it.
This isn't new. We've been doing it for a long time with booze, porn, drugs, sexual excess, gambling, pointless consumerism, certain kinds of religious fervor, endless things.
But almost all of those things are self-limiting. They're either costly, dangerous, in limited supply, or physically harmful enough to our health that we shy away from them and taboos develop around them.
Addictive digital media may actually be more dangerous than those things precisely because it is cheap, always available, endless, and physically harmless. As a result it has no built-in mechanism that limits it. We can scroll and scroll and chase social media feedback loops forever until we die.
AI slop feeds are going to supercharge this even more. Instead of human creators we will have AI models that can work off immediate engagement feedback and fine tune themselves for each individual user in real time. I'm quite certain all the antisocial media companies are working on this right now. Won't be long before they start explicitly removing human creators from the loop and just generating endless customized chum with ad placement embedded into it.
Some people have the discipline to push back, but many do not either for psychological/neurological reasons or because they are exhausted and stressed and unable to summon the energy. Humans do not have infinite willpower. So I've been predicting for a while that eventually we're going to heavily regulate or tax this space.
This concerns me too due to the free speech implications and the general risk of overreacting and overcorrecting. It'll be tempting for politicians to regulate or tax only the platforms they don't like, or to use the regulatory mechanism to crack down on legitimate speech by grouping it in with addictive chum. We've seen similar things with attempts to regulate porn or hate speech. But it's coming. I have little doubt. I think we'll see this when GenZ and GenA start entering politics.
It's really still shocking to me. If you went back in time and told me in, say, 2006, that our engagement-hacking would be so successful that it became an X-risk to humanity, I'm not sure I'd believe you. I never would have believed how effective this stuff could be. It's just a damn screen for god's sake! I think a lot of people are still in denial about this problem because it seems so absurd that a touch screen can addict people as well as fentanyl, but it's true. I see it around me all the time.
Edit:
My preferred way to go about reeling this back in would be to strike at the root and start taxing advertising the way we tax booze, drugs, gambling, and other vices. Advertising revenue is the trunk of this tree. The entire reason these systems are created is to keep people staring so ads can be pushed at them. Take that away and a lot of the motive to build and run these things goes away.
Another, which we're already seeing, is to age-restrict antisocial media. Young minds are particularly vulnerable to these tactics, more so than adults, and all addiction pushers try to addict people early.
Lastly, we could start campaigns to educate people. We need schools teaching classes explaining to kids how these systems addict and manipulate them and why, and public PSAs to the same effect. It needs to be treated like a health issue because it is.
Taxes, education, and age restriction is how we almost killed cigarettes in the USA, so there is precedent for these three things together working.
We also need to be a lot more precise in our language. The problem is not the Internet, phones, computers, "tech," AI, etc. The problem is engineering systems for engagement, specifically. If you are trying to design a system to keep people staring at a screen (or other interface) for as much time as possible, you are hurting people. What you're doing is in the same category as what the Sackler family did with oxycontin. Engagement engineering is a predatory destructive practice and the people who do it are predators. I think it's taken a long time for people to realize this because, again, it's just a damn screen! It's shocking that this is so effective that we need to have this societal conversation.
I don’t have anything to add, but just wanted to thank you for this insightful and deeply thought out response. The solutions you list do look like they would work and I hope we find the political will, sooner rather than later.
Patrick Boyle eventually comes to a similar conclusion in his video about global population decline - https://youtu.be/ispyUPqqL1c?si=7jUgVBkOvLHluPAR - but includes lots of graphs and other interesting factlets.
* warning for Americans: not suitable for those offended by sarcasm
I agree that advertising is the root of it, but some people might still pay for modern social media. They used to pay for porn, before it was available for "free" (ad supported). Some still do. I pay for YouTube to avoid the ads. I don't think I would pay for Facebook or Tiktok though. Possibly an uninformed opinion as I've never used those platforms.
Paid media is better. If you pay for a monthly subscription or for individual pieces of media, revenue is no longer tied to “engagement.” If you pay for Netflix and watch three hours a month or thirty, it makes little difference.
Ads tie revenue directly to time spent on the screen, and that is the root of all evil.
As another poster mentioned, ad revenue is often higher than what you can reasonably get with subscriptions. That’s where taxing advertising would help.
There's no way to reel it back. You said it here though:
> Some people have the discipline to push back, but many do not
This is simply a genetic selective filter that will destroy some people while others make it through, and there will need to be an overall adaptation against finding fake slop debilitatingly addictive. Like drugs, alcohol, porn, food, opiates, etc and other things some can resist and are able to abstain while some can't. I used to worry so much about these things in aggregate but I realized it's too pervasive to eliminate and impossible to change people's nature when it comes to resisting it or even worrying about it as a problem to avoid, so simply resisting better than others and having children that hopefully are able to overcome and avoid by way of finding more value in real experiences is the only successful outcome.
If one has to really really think hard about and try really really hard to overcome, then they're probably just not going to make it... and we all know for many people avoiding addictions comes easy. This chasm of reaction to stimulus means there will be divergent outcomes. It can't be any other way.
This is provably wrong. Preventative public health measures against for instance cigarettes and nicotine reduce uptake, reduce consumption and increase quitting. [1] In the case of smoking, this also cut second-hand harm/death from smoking. Similarly, preventative measures have first order and second order benefits for alcohol and other drug consumption.
Just giving up on those who show higher likelihood for addiction is a travesty. Failure to eliminate an addiction is no reason to give up reducing its harms, both to the person themselves, family and friends, and wider society.
It appears that the current content systems have some correlation in lowering the fertility rate; in that case they will be self-limiting after all, just not in the way OP mentioned about the other vices. It will be interesting indeed how things look after a generation or two.
This seems ridiculously fatalistic and weirdly binary way of looking at things. Best I can start with is 'why?', because to a simple person like me it could be any number of ways..
Eventually the curves cross. Eventually the computer you can get for, say, $2000, becomes able to run the best models in existence.
The only way this doesn’t happen is if models do not asymptote or if computers stop getting cheaper per unit compute and storage.
This wouldn’t mean everyone would actually do this. Only sophisticated or privacy conscious people would. But what it would mean is that AI is cheap and commodity and there is no moat in just making or running models or in owning the best infrastructure for them.
reply