Native transparency isn’t so hard to do by the way, I made an image AE (I don’t say VAE deliberately as none of these are VAEs, I don’t know why they keep being called that since the variational part is completely absent) that supported this about two years ago as a hobby project. I haven’t really been following the space recently, I’m surprised it’s taken so long for this to come out if it’s a first.
It’s the message board(s) that the OpenAI agent swarm was able to communicate through via GET requests. I guess a bunch of vibe coded weekend projects based around this idea have now dropped.
The absolute value of those numbers matters a lot. The cloud providers could be 100 times cheaper than running locally, but if it still costs say, 10 cents a day to run locally, you’re not going to care about this difference very much. And what you keep in privacy out-weighs the trivial savings afforded by the cloud provider.
I never said local models will disappear. There will be equilibrium. But excluding special applications where communicating with external servers is not an option, cloud is always going to be able to provide better inference for lower costs. That's structural.
> The cloud providers could be 100 times cheaper than running locally, but if it still costs say, 10 cents a day to run locally, you’re not going to care about this difference very much
For ad-hoc use, maybe not - but anyone running a business that's some form of pushing input through LLM to get output, will see costs proportional to use and error rate inversely proportional to quality, and they'll not be looking at it as "$0.1 isn't much", but "cloud lets me reduce costs 100x", and translate that to some mix of more volume, higher quality, and broader reach.
> And what you keep in privacy out-weighs the trivial savings afforded by the cloud provider.
That's even more niche than running LLMs on Martian robots. Most real privacy concerns are solved with contracts and audits. Individual ad-hoc use may lean more heavily towards local processing, but that's still a rounding error in overall use.
Yes, I think this is exactly right. The labelling of a thing as “AI Generated” or “AI Assisted” makes the viewer perceive it as worse in quality. And they really do think it’s worse.
It’s similar to a well known finding about price: people value things that are more expensive more, and tend to think they are of higher quality. You can give people the same thing for a high price or for free and they’ll say the high priced one is amazing the free one is lacklustre.
It’s like an optical illusion driven by labelling.
By many people’s criteria it automatically is worse, because part of what they value in art is thinking about the artist and their intentions behind their creation.
Similarly people may view an IKEA book stand as worse quality than one built by hand even if the hand built one has flaws. Or they might view a microwaved meal as worse than that created by a home chef.
People value the human intention and care that goes into creations. Without that it’s worse even if the output is fundamentally the same.
A drawing by my young niece may not have the same finesse as an AI generated one but I like it a whole lot more
>By many people’s criteria it automatically is worse, because part of what they value in art is thinking about the artist and their intentions behind their creation.
I think that's true enough for what it is but, as the other commenter also noted, a bit of a different topic here. The Monet example shows that people work backward from AI to direct conclusions about the quality of the work itself, which isn't the same as agreeing a work is by all appearances equivalent to Monet but lacking in underlying deliberation.
One can inform the other to be sure but that conversation exists in parallel with another conversation that's a straight one-shot assessment of quality.
> By many people’s criteria it automatically is worse, because part of what they value in art is thinking about the artist and their intentions behind their creation
That’s a bit separate to what I’m talking about, but I don’t think there are many of these people. I think this whole thing is about AI anxiety coupled with American social norms, and as soon as one of those things changes the artists are going to be left out in the cold again.
I think there are lots of people who value art for the ideas of the artist. There are also lots of people who think that they value something like quality of brushstrokes, and these are the same people, mostly. I mean they don't know what they like it for, but the raw graphical data is what's in front of them, so they ascribe their preference to its (supposedly) measurable qualities.
Hence this thing about I can tell it's AI because its graphical quality is bad. In a way that's correct though! The artist's ideas are hinted at in the brushstrokes (or whatever other graphical aspects). In the gen AI picture you get hints of a phony mind. It becomes hollow when you know that, or when you think you know that. It's not delivering the cultural experience, if the hints you pick up on are false.
There’s the idea of preferences and revealed preferences. People say their preferences and then their actions show what they really prefer. I suspect AI generated content, if curated by a human, would end up being people’s revealed preferences. There are a lot of bad artists and writers and whatever out there, AI would handily beat them.
I think at least for your niece’s drawings example, you’d probably prefer AI art if you had to choose one or the other until the end of time, not just one or two drawings that go on your refrigerator.
But of course with the gen-AI stuff you are getting put in touch with the real ideas of real artists (e.g. Miyazaki), it's just that they've been anonymized and put through a blender.
Revealed preference is a very real and useful phenomenon. But it seems to be increasingly overapplied as a convenient cop-out in scenarios where it's unclear any authentic preference is being revealed (e.g. due to confounding factors or active manipulation) and to exaggerate the extent to which real agency exists (externally predetermined outcomes).
Such as with RLHF-boosted writing from LLMs, where side by side comparisons of human vs. AI writing have been used as proof of revealed preference for AI. But an A/B test of a few sentences doesn't mean that someone would enjoy reading a 5000 word blog post in the same style, as the densely packed rhetorical flourishes become exhausting and repetitive tells start to jump off the page. Nor does it capture the signal/noise problem where despite technical ability and incredible verbosity, realization sets in that nothing in 2000 words has said anything at all while continually stringing the reader along. A side by side comparison with a few AI posters might be preferred in a five 5 minute test, but seeing the same visual elements from three frontier models in a loop on every surface for your entire life could quickly become actively repulsive.
And the problem with agency arises when limited freedom exists to truly make a different choice because the available options are externally determined. I've seen the argument made on HN that people's revealed preference is for AI maximalism, pointing to the fact that consumers keep buying products with AI, hundreds of billions of dollars are spent building new datacenters every year, workers (including creatives!) and students report using AI for daily work, etc.
But consumers aren't really given true choices that could reveal a real preference. The tech giants behind the phone duopoly have crammed AI features into both OS choices. With few exceptions every large company made an "AI strategy" their main focus since 2023 and competed to put chatbots and GenAI features into every inch of their product lines. Hyperscalers started a bidding war on datacenter buildouts in the expectation of future demand back when really only ChatGPT had any real market penetration outside the tech industry. Students and workers feel pressure to keep up with their peers by using AI to avoid getting fired/worse grades, artists struggling for work worry clients will go elsewhere if they take the time for their normal process.
Revealed preference when used haphazardly could be used to come to absurd conclusions like "people want to get scammed", "people like to be overweight", etc. Broadly and incautiously applied it just becomes a way to justify any given status quo as inherently optimally and reflective of collective will. With a creepy paternalistic implication that regardless of what you think you believe, forcing you do something against your will is secretly what you really want, and will make you happy even if you pretend not to realize it.
It's pretty questionable if a preference truly exists as "revealed" for something like AI art when it's preconditioned on active deception. An AI art-hater who gets tricked into picking AI doesn't mean they're lying/confused and actually love it. That sort of reasoning can get uncomfortably close to an anti-democratic argument for mass manipulation of people who can't be trusted to advocate in their own best interests.
Yeah, but the issue is we are getting constantly exposed to a barrage of really low quality obvious AI generated content. Except a lot of people seem to not know that it's AI and interact with like it's real. The people posting it are being reward by monetization platforms.
Low quality baiting content was a problem before but the barrier to entry has been significantly reduced and the.amount of content is growing quickly.
The broken window fallacy is the idea that it's good when your window gets broken because you have to pay someone to fix it and it increases their income. If that were true, then it would be bad if you could easily repair windows without spending time or money. It's false because we're better off if we can spend our resources on things we actually want rather than constantly maintaining our windows, even if that means some window repairmen need to find other jobs.
What do we do when former window repairman can't find other jobs? Just fuck em, right?
The narrative we're seeing around AI and current automation isn't "we're going to automate one category of jobs". If it were then sure, we could be relatively confident that the excess labor from that industry would be absorbed by other segments
But that's not what the future looks like to me. The future to me looks like "we want to automate a large majority of labor"
So what then? If tons of labour is displaced then the Broken Window Fallacy doesn't really apply, because it is based on the assumption that there will always be something for displaced labour to do.
Call it the Excess Labor Corollary. If there's no way to absorb excess labor in other segments, then what.
When there really aren't any jobs for the former window repair man to do, we are in a "post scarcity economy", which could be a utopia where everyone lives in comfort and has more leisure time.
Or of course, a few billionaires could hoard the wealth.
There's a Wikipedia page about it.
We could be in a post scarcity economy now, except that generally instead of realising we all have enough of we share, we raise the threshold of what we consider to be enough comfort and stuff.
>could be a utopia where everyone lives in comfort and has more leisure time.
It definitely could be. With current trajectory as of now, do you think it will be?
>except that generally instead of realising we all have enough of we share, we raise the threshold of what we consider to be enough comfort and stuff.
I see the opposite sentiment. I see previous basic necessities further out or reach then they were last generation. People aren't raising their standards, they are sinking towards poverty.
But then their worries are dismissed because they can get a flatscreen TV for 200 dollars. A TV that is likely feeding your data back to corporate to make thousands off you.
You're going back one generation, and I'm thinking further back.
Like, if we all settled for a 1950s British living standard, using current manufacturing technology, we could achieve that while everyone works a little bit less.
But, we'd prefer to keep labouring and have nicer things.
The supposed goal of "Make America Great again" is to invoke that 50s era of American exceptionalism. So surely there's something there people miss.
And your theoretical implies we have choice in the matter. I'm sure many would in fact take a smaller home with less amenities over a modern home with cheap appliances. Not that I'd actually be given that choice given my family lineage, but the general living standard of the 50's was quite high. I doubt the act of AC and the lack of knowledge of abestos are the things that made housing skyrocket over the decades.
Sure, but if we run AI all the way to its endpoint, it’s not just about window repairmen anymore. Our current notions about economics assume there’s always another job class for people to move to, here eventually there won’t be.
So here you never have to pay anyone to fix your windows, because they fix themselves, but conversely nobody will give you money to fix their windows! And then you have to pay your self-fixing windows a rent in tokens without an income.
I don’t think so, even if the whole “AI exponential” idea is wrong, that’s just the shape of the curve. There still is a curve. Even if these things take 50 years to happen instead of 5 (which would be better and which I doubt will be the case) it still happens eventually. I think the current capabilities of AI are enough of a heuristic for us to assume this is going to happen.
It is a stated goal of several prominent AI business leaders. Altman, Amodei, and Musk at the very least have had mask-off moments where they said as much.
People aren’t feeling the benefits of AI. Where there are any benefits, they aren’t causing prices to decrease or supply to go up, instead the excess value is being absorbed by profit taking.
If most people aren’t shareholders but instead workers whose jobs are threatened by AI, we should expect this response.
It has nothing to do with artists, it has everything to do with trying to mount a collective response to the apparent threat, poorly expressed as crowds tend to do.
Interesting. Might explain why the cerebellum and cortex work so differently? I wonder what the two ancient nervous systems looked like at the point where they merged.
Its a joke. You know they say ADHD types made great hunters thousands of years ago, because they can hyperfocus on that gazelle while constantly keeping their head on a swivel.
But your quote isn’t a quote from that post, you’ve made a straw man. What if what he says is true? What if less intelligent people are more inclined to believe those who seem intelligent, but without beging able to gauge their intelligence correctly? Or the same with authority figures, or simply agreeing with “what everyone knows”? All of these are heuristics we should expect when reasoning is difficult for a person.
These are also not the only ways that people can be misled about things, and susceptibility to them is likely not exclusively correlated to intelligence, but they might well be heuristics that LLMs can leverage effectively.
Also that correlation I mention likely can be quantified, but I doubt anyone is doing the research on it.
>What if less intelligent people are more inclined to believe those who seem intelligent,
I'm going to assume the actual truth behind this is complex, varied, and social.
When I was younger I had moved from a school that slightly above average to one that was below average in grades. The area had a really strong cult of ignorance feeling and I saw a lot of different behaviors in people that reflect themselves in political patterns. People that embrace ignorance seem much less likely to follow people that outwardly signal intelligence, quite often the opposite happens and they outright reject them. Instead of intelligence they are much more apt to believe those that broadcast success.
This quite often leads to power seeking individuals "playing dumb" and bolstering their success signal. "Awe shucks, that guy that got top scores at Harvard is just like me who didn't graduate high school" would be a good analogy to frame it.
You mean you don't independently know one way or the other?
So you are taking it on his authority?
Because it was stated with confidence?
According to him, you must be one of those low IQ people.
Does that help to see the irony here, and why the original claim is not only clearly wrong but inherently elitist and insulting towards its own intended audience?
The alternative, the world that all intelligent people must live in at all times accordinng to the comment, is epistemic solipsism, that everything is presumed false, everyone is lying or erroneous, because there is no guaranteed way a priori to separate true statements from false ones.
> I doubt anyone is doing the research on it.
Not only is there research on it, this is like Psych/Journalism/Philosophy/Law/Communication 101 stuff. There are textbooks on it written over a century ago. It is the topic that the field of "Social Psychology" was invented to study.
The dominant theory is the Elaboration Likelihood Model, which describes and predicts based on situations when a person is more likely to doubt a claimed statement of fact, taking into account past experience with the source, cultural norms, motivation, time, concentration/nondistraction, and domain knowledge.
There is also tons of research specifically on the impact of intelligence on susceptability to persuasion by authoritative tone, and the reality is the *opposite* of what was claimed: smarter people who enjoy deep thinking are more likely to believe falsehoods (Price & Stone, 2004), because they perceive the overconfidence as exciting new information to think about, just like you are doing now.
So congrats, by uncritically believing the guy who incorrectly said you were low IQ for believing him, you actually provided evidence that suggests you are high IQ, because science shows that gullibleness is a trait of high-IQ people.
You could also be right-wing authoritarian, that's the other thing that strongly predicts gullibleness but I'm gonna guess you're smart instead.
Not trying to combat you, trying to combat the other guy, who is being elitist and condescending and also incorrect in very ironic ways.
I'm guessing you didn't read what I wrote, or maybe I flubbed the communication in some ways because I was trying to be cooperative and complimentary towards you.
Edit: ok, you might object to me saying you're gullible, but the point I'm making is that that's not a bad thing and is actually evidence in favor of intelligence, the opposite of what the other guy was saying.
I'm gullible too. I'm also short, nearsighted, bad at communication, unable to bench over 90 lbs, and tend to spend way too much time researching fields irrelevant to my own work. Susceptibility to over-confidently incorrect messaging is a scientifically proven achilles heel of smart people. It's important for us to know that so that we can metacognitively account for it and hopefully avoid falling victim to it by way of scams, cults, and hallucinating AIs.
reply