Hacker News Reader

Comment by mjr00 | Hacker News Reader

This isn't really surprising. Even the HN comments for ChatGPT/Stable Diffusion related topics completely drank the kool-aid for a bit. I saw absolutely unhinged blanket statements like "ChatGPT is going to replace junior developers;" "Stable Diffusion is going to destroy the art industry;" "in a few years people will listen to completely AI-generated music playlists on Spotify."

The hype has died down a bit and sanity has been mostly restored to the comment sections here, but holy hell, HN readers are in this industry, they should know better, yet so many were, and maybe are, still fully bought in.

The average VC, or other manifestation of "fool with money", has zero chance. The macro economy aside, it's absolutely going to be a feeding frenzy for the entrepreneurs and grifters who can put together a convincing presentation on how they're making an AI Netflix that will show infinitely generated AI TV shows or whatever.

View on HN

> "Stable Diffusion is going to destroy the art industry;"

While there is of course a bit of hyperbole in that statement, I do think it has a lot of truth in it as well.

Stable Diffusion isn't going to destroy art since art is fundamentally a human endeavor. But many of the creative professions commonly called "[something] artist", whose day-to-day work consists of drawing pictures to illustrate a point, making things "look good" etc. are absolutely going to be wiped out. Logo designs, cover illustrations for books and music albums, pictures for marketing releases and so on. No one is going to keep paying people to do those when AIs can do it faster, better, and at no cost. I wouldn't be surprised if already in 2023, AI had a noticeable financial impact on those professions.

Totally. We commonly employ "tech artists" to do fairly tedious things in avatar asset creation (we build roblox games for brands). This is slowly going away thanks to ML tools being built internally by roblox.

Idk if it’s going to replace the “artists that make logos” industry, for example. I’d wager that the venn diagram of people with businesses that want logos and people who know how to use stable diffusion to get what they want is going to be exceedingly small. SD is just going to become a new tool in the artist’s tool belt that levels the playing field for geeks and allows them to encroach into the world of art. It’s just a new kind of artist.

You pay an artist to make decisions for you that you have no expertise to make yourself. Stable Diffusion and similar things don't help with that because ultimately what it requires is taste. The job might change and different people might be better at it than current artists, but you still need a human making that decision.

That isn't how commercial art works in practice. Executives rarely turn art decisions over to artists. What happens is that artists come up with multiple proposals, and executives pick the one they like best, usually after several iterations of suggestions from the same executives have already been implemented. Generative AI fits that process like a glove.

Yes, I know that, but where we disagree is that I think the social process of involving the artist is important and can't be replaced by having an app just show a bunch of options.

Lots of corporate art is already being created via outsourcing platforms that tightly control communication between customer and artist, who in many cases never even speak on the phone, much less face-to-face. That process is not very different from what a well-implemented iterative AI prompt could offer.

I understand that when a company like Coca-Cola is looking to reinvent their brand identity or similar, things are a lot more involved. But 99% of commercial art projects aren't anywhere near as complicated as that.

> Stable Diffusion isn't going to destroy art since art is fundamentally a human endeavor. But many of the creative professions commonly called "[something] artist", whose day-to-day work consists of drawing pictures to illustrate a point, making things "look good" etc. are absolutely going to be wiped out. Logo designs, cover illustrations for books and music albums, pictures for marketing releases and so on.

Logo designs are absolutely critical for companies that want a consistent marketing, advertising and branding (so, like, all of them). AI is also notoriously bad at making legible text. Saying Stable Diffusion can design company logos is the marketing equivalent of saying you can use ChatGPT to design your company's backend database schema; only true in the most reductive and simple of cases.

I would look into the current process for cover designs for books and albums. Either the commissioner is requesting a drawn illustration in the artists' style specifically, in which case AI doesn't help; or, if they're requesting generic cover art, the process is already very optimized. Illustrators heavily use stock images, do postprocessing in Photoshop, put text on as appropriate, and call it a day. These illustrators will start using SD as part of their workflow; if anyone is a potential casualty of SD, it's going to be Gettyimages and their ilk.

Again, this is buying way too much into the hype, and is a typical perspective of someone with little direct experience in paid artwork. The comparison to blockchain is apt; many people were talking about how blockchain would solve problems with "supply chain", "voter fraud", etc because they thought about the problem for 10 minutes and it seemed good. Of course, none of those people had ever worked in those areas, but it didn't stop them from talking about how disruptive blockchain would be!

edit: To put it in perspective, Tropicana's infamous rebrand cost them $55 million[0]. That's how much of an investment companies put into branding exercises. You really think they're going to be satisfied with a "good enough" AI logo at that scale?

[0] https://www.mashed.com/923840/the-ill-fated-move-that-cost-t...

Its already happening. People here are just not in the industry. Of course its not replacing everyone but its replacing certain subset of that work. The way forward its gonna be about specialized datasets like midjourney that stole all work from artstation and now does tedious commercial art really well (and gets better all the time).

Artists will be forced to use these tools. Whats more they are also gonna be always in competition with them. This will bring their wages way down and over time they’ll loose any agency they had. From valued professionals that spent decades honing their skills to being run of the mill AI photobashers.

Not great if this is future of knowledge work professions.

I often contract with companies who pitch creative work, and one of the trends I’ve noticed is that creative directors are no longer getting their employees to create concept art for pitches. It’s all Midjourney now. My sample size so far is three companies, but I suspect it’s an industry-wide trend.

There's a big difference between NFTs and most DeFi, and LLMs and other generative AIs. NFTs and shitcoins clearly had no actual non-scam usecases - they didn't actually enable anything genuinely interesting to be done that was previously impossible (except for breaking the law in some slightly new and interesting ways). But LLMs and friends... these bring things that were impossible, or impractical, or unscalable, into the realm of possibility. There will be genuinely important things built on this tech. Figuring out what is going to be properly hard work though.

But of course, there will also be a ton of lazy ill-thought-out crap, too.

And the danger is that NFTs and cryptoscams created a bunch of random lottery winners who think they got rich off their genius insight into how to apply innovative tech, but who actually only proved they know how to generate ill-thought-out crap, very quickly, at the start of a hype curve. So unsurprising to see them moving in. But without the obvious get-rich-quick scam opportunities, and the likely massive compute fees doing anything interesting in this space will involve, it's less obvious how long those folks will stick around.

I've also seen plenty of serious journalism and comments about ChatGPT destroying Google search, writing, journalism, art, education, and on HN the usual bits about countless programmers thoughtlessly copying/pasting buggy code into production without testing/QA/getting fired after the 3rd time, or even people shoving pills down their throat because they read some health stuff on ChatGPT without second thought.

There's a lot of people getting hysterical about it... just like there's the usual true believer audience (see: new age/conspiracy/etc in the past and now crypto). Every major new change generates these extremes... and no its not unique to HN or crypto bros.

Personally I'm skeptical things are getting worse because we still have the same reputation and social critique systems to keep these groups in check.

But it's always good to keep on top of the current things these groups latch on to... including both doomers AND culty/conspiracy extremes.

Here’s the full picture. Since 2018 we have experienced a series of tremors in AI. We know that eventually the big quake will happen — it’s inevitable. Some people said the tremors were directly preceding the big one, most people denied the tremors were even happening. Now that the tremors are over, we have a new geography where things are different but thankfully chatgtp has turned out to be very advanced search. But that’s not all it is. And it’s clear that we are building the subconscious of what will eventually become AGI. People who are alarmed at these recent surges forward are justified. People who assume the worst are justified too. We can’t turn it back. And no matter when AGI comes, people will think that it has come too soon.

> Since 2018 we have experienced a series of tremors in AI. We know that eventually the big quake will happen — it’s inevitable.

You can say/type anything, including this. But you offer no actual argument why this is true.

> And it’s clear that we are building the subconscious of what will eventually become AGI.

No it is not “clear”.

This. It’s not clear and I see a lot of ways the story could go differently. If the past is a guide, another ai winter is much more likely than agi.

Intuition feels the same when it’s right and when it’s wrong and it’s not a guide to decide what is clear, let alone inevitable.

Show proof/evidence or accept that it’s just an unfounded opinion.

This is laughable. Proof? How could anyone prove that AGI is imminent in a mathematical way? You would need to know the mathematical underpinning of intelligence and nobody will have that until after the fact.

It’s more likely for a winter? Dude, you either think AGI is inevitable or not. I never said how many winters are between there and now. But the idea of a winter now is completely insane. What is it precisely about the past 6 months that indicates a coming winter?

Show proof or evidence. Lol. The evidence is clear as day. The indicators and heuristics. Look at what software did 20 years ago and look at what it does now you fool.

The throughline from what software did 20 years ago to what it does today is pretty clear.

There's absolutely no throughline from the current ML projects to an AGI. The kinds of things they're doing don't even resemble conscious or abstract thought. They are probabilistic engines trained to mimic one thing at a time, only responding to specific input prompts, with moderate success.

Just because you can say that the computational structures being used resemble those in a human brain doesn't mean it's capable of the same things.

This "AGI is unfalsifiable" "either you believe or you don't" stuff is religious zealotry, just with a different focus.

Nah. I don’t think you understand consciousness. You claim there’s no connection but you couldn’t describe how consciousness works let alone how the models work. Don’t put words in my mouth, I never talked about the architecture.

There is no throughline from software 20 years ago because software was written deliberately then. All the experts said we wouldn’t be anywhere near where we are today by today because of this. If you can’t see the paradigm has shifted then you are simply wrong.

Proof should be required to continue forward. We shouldn’t need proof to do what’s safer. Especially when the consequences are so extreme.

I understand consciousness well enough to name two necessary-but-insufficient prerequisites for it, which I call continuity of input and continuity of learning.

The former means that it has to be always-on, with some form of continuous input stream, like our senses, as distinct from the current ML systems that sit there idle, waiting for discrete input and giving discrete output.

The latter means it also needs to be, effectively, continuously using that input to re-train itself, as distinct from the current systems that are trained once or multiple times in discrete batches, then used over and over again with that static model.

HN actually is generally quite negative on AI/ML and are the first to point out that humans are vastly better than this so we have no worries. This is indeed generally true for people on HN.

However, it seems like many here never talk or see ‘other people’ as I have said and keep saying; if you use fiverr or upwork, for the vast crowd of workers (as they are called) are far worse than stable diffusion (compared to the ‘artists’) or chatgpt (compared to programmers, writers, reviewers etc) now. Not only in what they sell, but also their communication about it.

So this is now; soon these people have 0 chance if they don’t pivot.

To be clear, I still consider "AI" as nothing more than overgrown MS Word macros. I have seen precisely nothing to contend this perspective.

But going beyond that I've just felt something has been "off" about the whole AI fad in recent months. I wonder if this is the reason behind that dissonance; the absolute insincerity of the people now doing most of the talking.

>ChatGPT is going to replace junior developers

Funny enough, when chatgpt was released on the first day, I managed to get in and posted the interview questions we used to ask at Amazon, both the behavioral and coding, and forwarded responses to my past manager asking if he would hire this person. Manager said, and I quote, "Yes, guy sounds very competent"

Reply on news.ycombinator.com