There was a book by Roald Dahl called the great automatic grammatizer which described an automatic typewriter that could be tuned to automatically create novels. Mediocre but quickly producible content swamped the market and genuine authors found it hard to get paid and so turned to these automatons to ghost write for them. A few held out and the cost was starvation. The story ends with "give us the courage to watch our children starve".
It was prescient. And i think this article captures the same sentiment.
> mediocre but quickly producible content swamped the market
I hear stuff like this all the time. It feels odd, given that pulp fiction dimeback "novels" exist for over a century. Some of these series run into several thousand issues, with authors expected to write a new novel each week.
The "mediocre but quickly producible content" has existed for ever, and it hasn't drowned out the "serious" novel market. Why should it now?
I'm not sure if you're American but take a look at the food situation here. Eating out practically anywhere is total garbage food. If you want high quality stuff made by people who know what they're doing be prepared to pay big bucks. A little pastry shop opened by me with a professionally trained French chef. The croissants are out this work but are like 8 bucks a piece whereas the Costco ones are maybe a dollar a piece and taste good. The market obviously incentivizes consumers to purchase the Costco variety but that's b/c the Costco variety is being more heavily subsidized than people realize.
Seems weird to be writing about croissants here but the Costco croissants are no where near the quality of a croissant made by a skilled baker, to the degree it is hard to call them a croissant in anything but form. The ingredients, flavor, flakiness, number of layers, etc. are all inferior when compared to what is possible from an artisan. I've bought the Costco croissants a few times in the past to make sandwiches out of as that's about all they are good for.
Your $8 croissants are subsided just as much as the $1 ones at Costco. They are cheaper because they don't have to pay that professionally trained french chef.
If the subsidy is like 1 buck/gallon of milk, the expensive milk will stay expensive even after subsidies, while the cheap one will come up almost free.
Mass production is cheaper than mass customization is cheaper than artisanal or full custom production.
That’s true in the absence of subsidies and I’d wager that the relative or differential effect of subsidies is minor in terms of explaining why $1 croissants sell well at Costco.
> I'm not sure if you're American but take a look at the food situation here. Eating out practically anywhere is total garbage food.
I'm not sure how old you are but when I was growing up in America in the 70's, eating out meant either high-end/pricey restaurants or MacDonalds, Burger King, Pizza Hut, or Taco Bell. Now you can get Thai, Sushi, Indian, Sichuan, and more.
> I'm not sure how old you are but when I was growing up in America in the 70's, eating out meant either high-end/pricey restaurants or MacDonalds, Burger King, Pizza Hut, or Taco Bell.
I grew up in US flyover country and I remember the 70s and that's not the case.
There were diners. There were mom- and pop-owned holes in the wall. There were more respectable "family restaurants" that didn't serve what was called "ethnic food" at the time but were certainly not fast food and were also reasonably priced. I remember my 15k-ish population town's one Italian restaurant and its one Chinese restaurant and the short-lived Greek place, none of them chains but all highly Americanized. Half an hour away in a nearby small city there were many more options.
> Now you can get Thai, Sushi, Indian, Sichuan, and more.
Those are all things we can find in the present-day equivalent of a 70s food court. They're part of the new fast food - mostly garbage.
Don't think the word 'subsidized' is being used correctly here.
Nobody is paying Costco a 'subsidy' to produce cheap croissants.
The big box stores incentivize the supply chain to do things faster/cheaper. Cheaper ingredients, shorter baking times (hence maybe some other additives needed), etc...
So, cheaper butter, oil additives, cheaper flour, additives to streamline short cooking times, easy movement, texture for packaging, etc..
Wheat and dairy are both subsidized in the US, so it’s not incorrect to say that Costco’s croissants benefit from subsidies.
I think it is incorrect/misleading to suggest that they benefit substantially from subsidies that are not available to the $8 croissant maker. (Costco benefits more from highway subsidies, for example, but not by enough to move the needle between $8 and $1…)
Agree. But it is difficult to tease out how the huge Agriculture Subsidy impacts each product.
I'm 'assuming', that if it is something like 1 cent, or 5 cents. Then it is a larger impact on the 1$ Costco Croissant), versus the 8$ bespoke croissant.
Then we'd also have to look at quantity sold.
Crouasant Sold Price Per Unit Total Revenue Cost Of Goods per unit Subsidy Total Cost of Good Total Subsidy Profit Subsidy / Profit
Costco 100000 1 100000 0.2 0.1 10000 10000 90000 0.111111111
Bespoke 100 8 800 0.5 0.1 40 10 760 0.013157895
So, it is true, Costco makes a larger percentage of profit from subsidies than the bespoke.
11% versus 1.3%.
The Quantity is much bigger factor.
So guess, on balance, Costco is more subsidized than the small baker.
Man, we should include labor too.
Basically small bakers are at huge disadvantage.
Edit
My spreadsheet didn't look good here. Is there a way to format HN to show a table?
I mostly read non-fiction and won't try speaking to the novel market, but in other pop culture media, it seems what happens is more the long tail of the reasonably good, still has artistic value middle that hollows out. There's always going to be a high end, if for no other reason than sufficiently rich people are often willing to lose money patronizing great art for its own sake, but the vast bulk of new content becomes mass-produced generic trash with easily replaceable creatives who are given no real creative control. I'd say it happened to mass-market film somewhere between 2005 and 2010. It's probably happening to serial filmed content right now, thanks to the zero-interest streaming wars being over and peak TV coming to an end. It arguably happened to pop music as early as the 1980s, but music is comparatively cheap to produce and starving artists will always self-finance, just discovering their work is difficult when it's made outside of a recording industry marketing machine.
Novels seem like they should have the same dynamic. It's effectively free to write novels. You only pay with time and stress and don't need money. But without a publisher to market you, the hope that anyone will ever read what you write gets dimmer and dimmer.
In the same vein, I feel like serial cartoons are holding out much better right now than filmed television. They're still in a golden era, probably because they're so much cheaper to produce that they can survive not being popular without getting canceled.
The issue is that for a hundred years, even that 'mediocre' content took a bit of skill. Even trash authors spitting out a book a week, are better than most people at writing. Not everyone can write a book a week, no matter how trashy. So the volume of trash was still limited.
The new tools have dialed the trash volume to 11. It is now possible to have infinite trash. Thus the value goes to zero.
There are a million novels produced annually. Most of them don't sell 100 copies. We already have essentially infinite trash and the majority are worth essentially 0.
There's a marked difference in having an author spend a week crapping out a cheap pulp novel vs having a text generator spit out 5000 in the same time period. Coupled with self publishing and automated stores, the market can be flooded with pure garbage in hours.
There are a million novels now, of which 90% is self-published for a reason (no one wants to publish it, because it won't make any money). With a text generator, there are now 10min lion novels, of which 99% won't make any money.
The book market has found methods to deal with the "99% is garbage" problem. This isn't new.
I am curious why vouch for provenance rather than vouching for quality. In “mass-produced AI garbage” is the “AI” part a bigger problem than the “garbage” part?
> I am curious why vouch for provenance rather than vouching for quality.
Because the tech community only has ideas that involve solving problems with tech. Provenance is (relatively) easy to address with tech. Measuring quality might require human input, so that's right out.
I've sort of felt like this has already happened in many aspects of my life. To be an American consumer feels like living in a perpetual 4.5 star limbo these days (I note that nearly every Amazon product I look at now has a 4.5 star rating). Every experience seems to be optimized via algorithm (perhaps orchestrated by computer, or perhaps just via one human organization or other) to be as bland as possible. Mass entertainment seems the most obvious - AAA games have become money-printing slogs, and the latest blockbuster movies seem ever less original.
It's possible this is at least partially an effect of getting older, but I do tend to find smaller works to have a bit more character and chance to be entertaining. Obviously the hard part is sorting that out as the signal to noise gets smaller.
> A few days later I found myself in a grocery checkout line, skimming through yet another article in which the writer touted the wonders of “artificial intelligence” and fretted hazily over whether we are nearing the point when AI will be able to produce novels, films, and other creative work, effectively replacing us. When I looked up and over to other people in the line, half of them wore the same shoe brand as me.
This is an interesting phenomenon with AI. People immediately assume that because AI has the ability to flood the market with mediocre art/writing/whatever, that artists and writers are screwed. In my experience it's quite the opposite; when the market gets flooded with repetitive mediocre garbage, the relative scarcity of something unique increases and actually increases the price, accordingly.
To the author's point, you can get the same Adidas shoe as everyone else for $50, or you can get a largely similar-looking pair of Guccis for ~$800 US[0]. Those Guccis would not exist if not for the mass-market cheap option; there would be no reason to splurge on an expensive unique purchase if everyone else was already wearing something unique. Or not wearing shoes at all.
This has been my experience since the introduction of stable diffusion and dall-e as well. Just anecdotal, as I don't have any access to hard data, but the artists I've spoken to say that their commissions and fanbox subscriptions have gone up since SD was released. Much like the trope of friend groups keeping an ugly friend around to make themselves look better, the mediocre and soulless AI art that's flooding the internet has, apparently, made people appreciate human output as a premium brand.
> Those Guccis would not exist if not for the mass-market cheap option; there would be no reason to splurge on an expensive unique purchase if everyone else was already wearing something unique.
I would have a slightly different take. They all exist for the same reason: extremely cheap labour overseas. That's why there's enough spare cash sloshing around to provide what is the main differentiating factor: marketing. Get the marketing right and you can charge far more for almost the same product. If your shoes were made locally and cost $200 on average, you'd a) pick shoes for longevity, not looks, and b) not be that impressed with shoes that cost "only" 4x more.
To translate this to AI: they're both using AI. Just one has better marketing paid for by higher prices.
AI art isn't as good as human art. But in a few years it will probably be better. It's a foolish mistake to look at the state of AI today and say "Actually this is gonna be great for my creative business!"
What is scary about AI is not ChatGPT or Midjourney, Dalle, whatever. What is scary is that they showed up 10 years sooner than the earliest predictions, and are still rapidly improving. That is what should give people pause and it's amazing how many people miss this.
> AI art isn't as good as human art. But in a few years it will probably be better.
This is a very big "probably". About as big as the "in a few years, AI drivers will probably be better than human drivers" line from 2015.
Regardless: an AI can definitively be better than a human at something like Go (a closed game with a clear win condition) or driving (a logistical task with no creativity). But how do you define "better" when it comes to art? From a technical perspective, AI is already better than the vast majority of human artists! Very few artists would be able to, say, draw images of nature to create something that has the same contours as a popular meme[0].
So, arguably, we're already at the point where AI-generated art is "better" than human art. The question is, does that matter? Or does it matter in the same way that it matters that a motorcycle is faster than Usain Bolt?
By better, I mean the point at which having artistic ability is no longer a viable money maker. It already hardly is, but I know a few artists who get by working in corporate graphics departments doing all manner of graphic design.
There will come a day when someone realizes they don't need a team of artists, they just need one. And that one doesn't need to be that good (read: paid well) either.
Human art isn't going to disappear. Definitely not. But the value of it is on the cusp of falling dramatically. Banksy will stay popular and still command millions. However that kid who you graduated with that is "an incredible artist" probably just went from a 75% chance of a decent artistic career to a 3% chance.
Yes, this will shift markets a lot.
But hasn't this happened a lot of times?
There are almost no tailors anymore, but in the 1800 going to a tailor was the only option to get clothes. Now it's a niche.
Now this will happen with graphics and writing as well.
Handmade graphics will become a niche, and there will be top graphic artists working for new AI models, like top tailors now do in fashion.
This is just the beginning of extreme market disruptions.
All commercial driving will certainly be automated within the next 25 years.
Things will change. Markets will adapt.
And if the rich are smart, they won't let people starve. Because this does not serve anyone. Hungry angry mobs will sow chaos and destruction.
>showed up 10 years sooner than the earliest predictions
Bit of a nitpick but the best known predictor of this kind of stuff is probably Ray Kurzweil and he's predicted AI passing the Turing test for 2029 since 2005 or so, so the state of AI being a little pre Turing now is almost spot on.
He also got AI beating humans at chess correct to within a year. It's not actually that hard - you plot computer capabilities on a log graph and extend the line a bit.
What is maybe a little scary and amazing how many people dismiss is the bit after 2029.
Top 10 summer movies 1982-1989 - 25% are sequels. In the 2010's that ticked to over 50%[1]
This is way beyond AI. Culture just repeats itself, from movies to fashion to music. In case anyone was wondering - in the never ending loop between oversized baggy jeans and skinny jeans, we are right at the very end of the baggy era, looping back to skinny in a year or 2
So is this article complaining that instead of looping between baggy and skinny we should all be wearing silver spandex space suits? I guess thats true, but also Im good with just jeans personally
> Culture just repeats itself, from movies to fashion to music.
And including AI. If you watch movies, from around 90-95 you will notice the same talk about AI taking over the world, jobs, and all the schizofrenia surroinding it.
I agree with you 100%. Recently, I've been generating landscapes with Midjourney, signing my name at the bottom, and hanging them on my wall. It feels good to finally overcome the artistic challenges I faced in the past.
I disagree with you on your claim that the art has zero value - clearly the OP saw enough value in it to frame it and put it up on his wall - he saw more value in this than something he could have purchased because it was exactly as he wanted it. The fact that that art piece might have zero market value is irrelevant as the OP never claims to want to sell them.
I also fail to see why it is relevant to point out that the OP lacks artistic skill - it was the vision that was valuable to them; putting a paintbrush to a paper is simply a medium that they were able to circumvent with a simpler solution.
I use hammers and nails yet i lack skill and the goods i make lack value. I dont know what’s up with folks thinking that producing a one off using a procedural generator makes them artists and this point i am afraid to ask.
That's a horrible analogy. And you sound like a bitter person.
The person you commented created something he loved. He loved it so much he framed it. With your comment, you created nothing but negativity.
Seems like @dorkwood is the artist (creator), and you are the critic. :)
I think this is the reality that plays out longer term.
Large orgs and high level execs are seeing an opportunity to produce cheaper, lower quality product more quickly.
These same orgs seem largely blind to the fact that while they are working to make their staff redundant to surge profits, those same staff have been massively empowered with leverage that would allow them to build competing products that would be more competitive.
Work at an org as a backend engineer? Know enough domain to build a competing product, but have no design or frontend knowledge? AI to the rescue.
In tech the biggest threats to large orgs has always been small teams with drive and vision, those teams can be smaller now and likely made up of former employees.
I don’t mean this personally, but just because you feel or perhaps are even able to quantify amplified creativity and output, that doesn’t mean it’s not mediocre. After all, mediocrity is subjective. And even if you’re doing incredible things with it, there are most certainly a lot of folks (I’d argue more) who absolutely are not. In my senior role at a software company, I’m increasingly having to spend more time protecting our products, marketing, documentation, and support from developers, product and design staff, support department, and even vendors from jamming in broken, poor quality output from automation.
The article touches on a more philosophical (and some cases actual [1]) negative feedback loop: these tools are trained on human output, explodes into popularity, and–greatly boosted by seedier, low rent content farmers and the likes as well as mind boggling scale and speed–is fed back into the same input stream: akin to the “photocopy of a photocopy of a…” effect. I recall years ago scoffing at talks about how the Internet will soon be an unusable torrent of detritus that drowns out actual useful or interesting information and looks at Google search results here we are.
When it comes to generative tools, sure, there are defenses and mechanisms to mitigate this effect to a certain extent, but it also has a likewise effect on humans and society, which doesn’t, not really. We are constantly driving the bar lower, both from the business and consumer perspective.
A parallel: Look at mass produced, cheap consumer goods. I’m not ancient, but I’m old enough to watch my era of cheap goods (which my betters lamented over compared to “their time”) also become replaced by notably worse quality products than before. We don’t even seem to care if the thing meets some minimum threshold of functional in its purpose, it only matters that it’s immediately available and cheap enough to dispose of if it sucks especially bad. This is an excellent proposition for economies that have been working extremely hard to push society in this direction. It’s awful for everything else. This is a drain-circling race to the bottom with disastrous ecological and human implications.
Of course, you can dismiss this as shouting into the void, technological old man yells at the Cloud… which it is. The ram has touched the wall, Pandora’s box has been opened, the genie is out of the bottle. But when I compare it to other technological disruptions in human history, “AI” products and services are impinging upon widespread sociopolitical and economic stressors and challenges resulting from just the regular-ass Information Age. Since we are increasingly a technological society[2], and our most trusted but flawed quantifier for value and worth is money, profitability is the predominant motivator from big tech. Not so much its benefit to humanity or whatever delusion VC sociopaths and techno-narcissists are on about. AI is a tool. I’m not afraid of tools. I’m afraid of a race who has a difficult relationship with its tools and is prone to squander them on stupid or awful things.
I think the confluence of hyper capitalism, nascent technology, and human nature will make slowly boiled frogs of all of us, cut off from our true potential emotionally, intellectually, physically, artistically, morally, and ethically.
I think you(no offense) and many others look at this all as "mass production of artificial crap", HELLO all consumer shit is! I don't subscribe to that, so don't lump me in there.
Just like a milling machine can be used to produce wasteful widgets destined for consumer trashcans, so can the same machine be used to create high quality parts used in airplanes... my point is, YES, in the wrong hands any tool can be miss used.
You have no idea how or what Im doing with code gen, or statistics... and you paint it all "generative AI", yawn.
> I think you(no offense) and many others look at this all as "mass production of artificial crap", HELLO all consumer shit is!
My argument isn’t that “consumer rubbish is good as long as it wasn’t made with AI”, if that’s what you’re getting at.
Using your example, I am concerned that humans are losing what little ability they had to discern which widgets belong in the bin and which are high quality, that’s not an accident and AI is just one of many tools used by the wrong hands to that effect. To be clear, that isn’t AI’s fault, it’s a human problem.
I made no comment on your work, because quite frankly I don’t care. I agree that mediocrity isn’t some new fangled thing that appeared because of AI, but I think it’s pretty naive to dismiss its impact on the issue.
Edit: I just realized that I made a typo in my original comment that I apologize for because it most certainly did sound personal. Unfortunately I can no longer edit.
> The real danger to human creativity that these tools represent is the mechanization of human innovation. Relying on these tools will discourage us from looking beyond what has been done before, and further reduce innovation into no more than imitative remixing. Businesses and corporations assert that mechanization will benefit both the worker and the consumer. It’ll save time and make things more efficient.
Efficiency holds back innovation and creativity. It focuses on cost at the expense of discovery and exploration.
That quote: I don’t buy it. There is plenty of incremental innovation now and more ambitious people still shoot for the stars. If AI is mediocre and doesn’t solve what humans want to solve, it won’t replace human desire for real innovation.
Ideas are cheap and plentiful, everyone has them. They will be pursued, just in greater numbers.
Most of the incremental innovation I've seen from tech for the past ten years has been darkpatterns, enshittification, surveillance, and profit-driven exploitation.
No it doesn't. Efficiency unlocks bandwidth for us to explore on higher levels and new directions, and it always has. Efficiency of energy production, manufacturing, transportation, information sharing etc – you cannot seriously believe that these did not spark new waves of innovation and creativity.
Mediocrity is a good thing in the context of commodity goods. To use the author's example, you CAN still get unique custom made shoes and still buy hand-grown heirloom tomatoes at the farmer's market. The shoes will cost you $1,000+ and you'll pay a couple of bucks a tomato - prices which adjusted for inflation haven't changed since such practice was the only way people got shoes and tomatoes, and they got a lot less of them.
Mediocrity and mass-production have made shoes and tomatoes accessible to everyone at far lower cost, and are not inherently bad.
LLMs will predict the most likely output to the prompt input - does this mean that a hypothetical perfect LLM would produce the most generic and widely-acceptable response to any input, the "lowest common denominator" output?
The author seems to think so, and indeed, with today's GPTs, that is often what you might get back. But I think there's a lot more going on here - with natural language, a lot of the training effort goes into producing natural language output which conforms to the formal rules of language (it is intelligible, syntactically correct or roughly so, follows grammar, and semantically "about" what it is supposed to be about). But the interesting thing about natural language is that there are infinitely many correct responses which are valid formal responses to the prompt, both ones that introduce new or rare creative possibilities and ones that tread the same ground as a wikipedia article, for example (which is usually intended to be very neutral). So an LLM can produce both types of responses - formally correct and boring, formally correct and interesting - and you can see this today by offering "mix-ins" to the prompt like, "Tell me about Saudi Arabia's ruling family, in the style of Peter Griffin". Of course, it won't select to use the style of Peter Griffin without that prompt, but I think you can imagine a "weirdo" LLM that is trained with some positive response toward creative or interesting output too.
LLMs don't generally predict the most likely output, except when people go in to tweak temperature and other settings to make them more predictable for certain more mechanical tasks. The whole point of an LLM is that its output is calibrated; it produces a text with the same conditional probability you would expect seeing that text in the wild (give or take that we still don't understand why neural networks generalize).
Would the output with the same conditional probability not also be the most likely in a conventional sense? I'm not understanding the distinction, but I'm not an expert so this might be a subtlety I just don't grasp
Conditional probabilities are probabilistic. If you ask again you might get a different answer. The most likely answer is a single sample. It often isn't actually the result you want, it often has a much lower probability of occurring than your intuition would suggest, and even if it's correct once it's often not something you would appreciate appearing repeatedly if you re-asked the same question.
I can add more details/comparisons/nuance/explanations/... if you're curious why that matters, but hopefully the distinction is clear now?
I don't mean to imply that creativity is something that can be quantified objectively (used as the objective function, even), but it is widely understood that LLMs can produce varied output and be given particular biases or personalities
I've posted about this before, but this is just the hollowing out of the "middle" of the value curve.
Most people want something that's good enough, and they're not willing to pay for something that's N+1 good at 2N price. Other people want something that's N+5 good and will pay 3N price. As AI pushes down the lower bound of cost for creativity, the "mediocre" content creators/programmers/writers are going to get squeezed. At the same time, this means there is an untapped market for the people who can truly be creative and do great work.
Thinking the difference between the Sports Illustrated AI articles and premium sports journalism (The Athletic), blog spam recipes vs America's Test Kitchen, or CNN/Fox front page vs NYT/WaPo/Politico, etc. You're going to have to vote with your dollar to support the things you care about. If you're not willing to pay for the higher quality, you're going to get AI (or other "budget") content that's making it's money by just being good enough to get you to see the advertisements.
It's a free market out there and you need to be aware that just a few percent of people (probably roughly similar to the lurker/poster ratios) are the ones that enable high quality content to be successful. The HackerNews crowd are significantly more likely to be able to actively support creative pursuits that they find valuable, so I'd suggest you do so! Otherwise, when we don't reward quality, why wouldn't people just be mediocre?
It's maybe hard to fathom for individualists, but it's probably humans that are quite mediocre on average. These AI models simply represent that.
And on the value of generative AI, the same argument against new trends has been made on previous inventions. For decades, a typewriter was still more practical than using a word processor on a computer. For more than a decade, calls and SMS were more practical than online messengers and forums.
"The problem for AI is that creative work is not predictable. It is not about statistical likelihood or simply mashing up the familiar—it is about leaps in logic and counterintuitive juxtapositions. It is about the unique experience of the individual, and seeking to do what has never been done before. It is about the least predictable next word or pixel. So the danger is not that AI programs will write the next great novel or create the next great painting, successfully replacing human inventiveness: they never will. The greater danger is that they won’t need to create great writing or art."
I never get this argument. There is no magic in the human creative process. One way or another, algorithms should be able to replicate it and eventually surpass it.
I think the human creative process is more interesting than AI. I love learning about the person behind the creation, and hearing how their upbringing and experiences shaped what they created.
This is the most important comment in the thread. Humans want to connect with other humans. You can pump out all of the perfectly-rendered perfectly-customized & personalized fully-immersive metaverse simulation but at the end of the day, we bags of meat and gas will still want to sit around a fire and tell ghost stories.
The market for 'fully tech driven' content exists, sure, but it'll always just be that.
The examples I always think of are Kurt Vonnegut and Raymond Carver.
Vonnegut's experience in WWII heavily impacted his writing and views, and led him to create one of his best works.
Additionally, when I talked to a war veteran about Slaughterhouse Five he said that the time travel was very relatable. When he came home, he felt like he was moving backward and forward in time. Everything at his home was the same, and it was like he never left. It was such an interesting conversation, and something that made me appreciate Vonnegut so much more.
For Carver, I just really enjoyed learning about his life. His work experience, personal relationships, and experience with alcoholism shows up in almost all of his stories. I find it cool to read a story, and draw parallels to his life.
Additionally, my father was a blue collar worker and worked in many different jobs his entire life. I draw so many parallels from the stories he's told me to Raymond Carver's stories.
I really love forming connections like this to the art that I enjoy.
This is why there will still be a market for human-created art, but it'll likely be very niche, for people who care. When it comes to sheer quality, the AI should be able to surpass us at some point and create masterpieces we can't even think of.
It's more God In The Gaps. Something is declared ineffable, beyond the grasp of the mind, science, tools, AI. Then we have an undeniable breakthrough, and they roll back that stance the very smallest amount to accommodate the new knowledge.
There's a strong historical pattern that the appearance of the limits of reason has always been due to a lack of imagination, often in the very people who just spent their lives expanding its borders. And these arguments so often rest on "je ne sais quois," it comes off as more ridiculous than parsimonious. If you think you found an illogical or indeterminate system, I'd bet that your body of kmowledge just needs reframing.
Though ever further we might see, surely dragons further be! c:
Isn't the reverse of God In The Gaps just as ridiculous? Just because we have been able to use a materialistic, scientific approach to get this far doesn't necessarily mean it will keep working forever on everything.
Using your metaphor at the end: "There's dragons just over the next hill" and "There haven't been dragons so far so there will never be dragons" are both just guesses.
Creativity is not just cognition and pattern matching. There is something more fundamental at work... obviously calling it 'magic' doesn't describe it, but some people call it 'inspiration' and they feel it on a physiological level. They starve and torture themselves to serve it. To recreate something like that you would need to simulate not just a human's biology but an entire world.
I don't know. It's like Frankenstein's monster reading The Sorrows of Young Werther. He could somewhat understand what the words meant, but he couldn't ever feel what was being described. Even if the monster could write about human feelings, it would be just as an observer, a reader, repeating and copying.
"They" refers to the programs that exist (or more broadly anything based on "statistical likelihood or mashing up the familiar", which is the overwhelmingly dominant sort); "never will" refers to "never will write the next great novel or create the next great painting", not "never will exist."
I think the author might agree that they never will exist, but because of his actual point about what they will do to creative work and communities, not any physical limitations.
I think it is still a solid argument, especially when juxtaposed with anything related to developing something new and innovative, whether in the domain of arts, science, or business. But what I found insightful in this quote was the note about the least predictable things.
This implies you know how the human creative process works. I doubt you do: it's an unsolved problem, like most things in this category (intelligence, creativity) dealing with how our brains function.
The magic is the biology. Computers are not going to understand the pain of taking a bullet or childbirth. What can they write about these topics except trite regurgitation?
And yet when I watch the flood of movies, television shows, novels, music, and fine art available to me, I have a really hard time finding anything unpredictable. Great skill is involved in creating them (sometimes), but for every one experience that amazes me I have to slog through hundreds that really don't. At best I'm entertained. The output is formulaic, as you would expect from businesses guided by bookkeepers. The sad fact is that a large number of these productions very well might have been generated using an advanced LLM.
Yes, this is because Hollywood - like tech – is a subset of business, which optimizes for profits over creativity.
Which is why, as a creative, I'm not super-worried about AI destroying creativity. It might actually help democratize much of the tools one needs to make a movie or a book, but at the end of the day creative consumption is a human-to-human demand.
> And yet when I watch the flood of movies, television shows, novels, music, and fine art available to me, I have a really hard time finding anything unpredictable.
Wouldn't that just testify that Hollywood et al. are no longer really "innovating"?
Absolutely. They're going for all the safe, boring choices, and depending on the business side of their industry to make money. You don't show a profit by making the next Children of Men, you show a profit by tweaking the Fast&Furious formula and strong-arming theaters into showing it.
Hm, my own intuition is that it'll raise the baseline experience but then cause hyper-competition by human creators to exploit its powers to the limits while infusing further human creativity on top of what is mechanistically produced.
I do not think this will result in overall mediocrity. It'll lift those with no access to art or way to pay for human creativity away from absolute poverty, while allowing those that already had access to these things a more competitive market.
AI may have the opposite effect to the "five cars identical in make, model, year, and color", "same shoe brand" stuff in that you'll be able to have it design something unique for you. Already you can do that with digital artwork and it may extend to customising physical objects soon.
That article was interesting but holy crap what was the deal with the constant shifting of the layout. Ads kept popping in and out shifting the content up and down. Its unreadable.
I don’t buy it. Humans are also a prediction machines. We continuously predict next thoughts, feelings, physical motion etc.
I think what llms lack right now is:
1. Continuous input from through different senses
2. Ability to do experiments
3. More modalities like touch, sound etc
Once these capabilities are added then they will be comparable to human “creativity”
Tech (which is nothing more than a lackey of Big Business) will never understand that real demand for cultural artifacts – art, books, film – comes from the consumers' need for real human connection. (The rare exception might be a guy like Steve Jobs).
This is a huge advantage to artists and creative people because they can and will bypass that mindset entirely, and can do so with the confidence that they're connected to what human beings – consumers – actually want.
It's why 'tech forward' orgs like NFLX can't really do much except churn out derivative works, meanwhile you've got the Davids, the Wachowskis, the Cohens, Tarantino, Chris Nolan, etc. constantly pushing the limits of creativity and writing their own ticket, without the permission of big business and its lackey tech bros. When these guys 'break out' of course everyone wants to be their best friend, but in no way was Warner Brothers or Paramount able to mechanize or catalyze that creative genius.
What chaps my ass the most about the current crop of GAI is that its gone after writers and illustrators and artists because, in the process of sharing their works with the world, had that work hoovered up and used for training data without their permission or recompense.
Luckily, all of the mediocrity these systems shit out is obviously superficial and soulless. Yeah you might some day be able to do a fully reactive murder mystery in Zuck's metaverse, but eventually you're going to want to be intrigued and delighted by something truly fresh and human – and that can only be created by other humans.
Are there gonna be enough hats for all these people to eat in a few years?
I think the perpetual image in media that humans are a unique entity with a magical inner force has gotten to everyone's head. It's just a fiction guys. It strokes egos to sell copies. We are meat computers outputting computation. There is nothing special inside. On our current trajectory in a few years AI will be far more creative than any human. Then we'll have to trudge through grumblings about how it feels so "cold and inhuman" despite blind studies showing people greatly prefer AI works.
I think there is something special. Some special coupling of emotional states between us (manifests as empathy, feeling of community, feeling of connection). Not saying that this connection cannot be replicated or faked, this remains an open question.
Perhaps building an AI vastly smarter than us will conflict with replicating this coupling. Perhaps we will prefer artwork from an AI with which we do feel this type of connection. Perhaps this will make us build an AI that is really close to us in terms of experiencing feelings an emotional states, even if it means it's less "smart". This wouldn't be a dull outcome...
It was prescient. And i think this article captures the same sentiment.