This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
I don't know if it's because I skip most AI conversations, but I think you guys might have failed to discuss some things that have been happening for the last several weeks.
Reverse engineering with AI has gotten really, really good, somehow. It's so good that you can reverse engineer Microsoft Office and host it online for free. It's so good that you can submit your tokens to the Machine God and get World of Warcraft open source. You can now play the original Xbox Halo, but with 32 player coop (seemingly most of the players being Mexican). On your Nintendo 64. In fact, you can play Halo in your browser now. You can play lots of stuff in your browser now. I looked up an old game called Apeiron for old Macintosh, released in 1990 something, just a Centipede clone developed by Ambrosia Software, long since abandonware. There are two ports of it that I found in comments sections on Macintosh Garden. The classic decompilation site is shutting down due to sheer volume. And yes, there's also mods. Flying my Banshee across the Mojave almost makes me wish for a nuclear winter. My Foundry in Warcraft 3 just told me that my AWP research has completed, so I went on a tear across Lordaeron.
On the other side of this, AI can now shut down 70% of the Internet. This bug was in PHP for a long time, and only now discovered, because someone pointed a technological superweapon at it.
What the hell even happens from all this? Can Apple really continue to have a closed ecosystem? Are we going to wake up with no money in our bank accounts? Does proprietary code even exist anymore? How is Steam going to cope with having tens of thousands more games submitted to it? And why now? AI was good for several months up to now. Was it the release of tools that work with agents effectively? This seems a bit more significant than hundreds of mathematical proofs being published at once to me, especially because you can verify it works yourself, on your machine, and see the effects on the world right in front of you.
The economy and way-of-life of the first world as we know it might be ending. But at least we can play Bloodborne natively on PC.
The fears of AI superpowering the script kiddies were overblown, at least so far: https://eliastaylor.com/noexploit/
That's not to say that the exploit isn't bad, but it may have been overhyped. The actual "hack any website instantly" button remains in the hands of only a few elite hackers, if it exists at all.
This is bullshit. According to anyone who actually claims to have used these vibe coded clones, they are buggy and unusable. They are created by grifters who want attention and do not even use the tools they are cloning.
I unfortunately have to admit that there are some (not many) quite useful vibe coded apps starting to roll out, but this isn't it. As far as I know, all of the useful ones were created to solve some of the creator's needs, rather than just built to be something flashy.
I use the vibecoded photoshop clone. I don't know how it compares to actual Photoshop because I'm not deep enough into graphics to justify paying Adobe, but it sure beats Gimp for my use cases. Call it damning with faint praise if you want, but a small army of driven and intelligent people worked on that project for decades.
If I ever get the time (a big if), I'd like to install the clone on a VM and see how it compares. I say this as someone who pays for Photoshop, but who isn't a graphic designer or using it for anything work-related. The problem with a lot of this testing is that, even if you do have the full version of Photoshop for comparison, the relative quality isn't going to be apparent from testing alone. You actually have to rely on the thing in question for some time and see how it goes, and see how the frustrations build. To give an example, ChatGTP's image generation is actually fairly good at doing things that Photoshop can't, notably upscaling low-resolution images. The first few times I tried it, the results were impressive, and I started making it part of my PS workflow for stubborn images that would take too long to edit the traditional way. The problem (if you can call it that) is that it only works about half the time, and that's being generous. It's not merely editing the image but generating an entirely new one based on what it thinks the old one should look like. When it gets that wrong, which is often, it generates something that is not only an incorrect representation but one that can't easily be rectified in Photoshop. It also works better if the images are in good shape going in, so there's a lot of preprocessing involved.
The upshot is that, a not insignificant percentage of the time, AI actually increases the amount of work involved. To be clear, I'm using this primarily to edit album covers for which clean versions don't exist, and the work involved can range from minor editing of a decent-sized image to working magic so a small, poorly taken one looks presentable, to completely reconstructing an image from scratch using the original as a guide. I now need to add a brief extra step of deciding whether it's worth attempting to use AI. Then I need to find an image to use, and prepare it. The extra work involved comes from the fact that the best image to use for an AI reconstruction isn't necessarily the easiest one to edit manually. I would also add that even successful AI generations have a tendency to oversaturate the colors or otherwise screw them up, so there's an extra step of loading the finished image into Photoshop and dialing things back (I have to do this anyway because ChatGTP generates 1200×1200 images and I want 1000×1000). When it works well, the results are nonpareil. But getting it to work well means doing 75% of the work for a 50/50 shot at something usable. It isn't so much a time-saver as it is an overall value-add.
The whole experience kind of sums up my opinion of AI workflows in general and why I'm not convinced that it will dent employment in any significant numbers. It seems impressive when you first use it, but when you actually try to implement it, it doesn't produce results with the consistency one can expect from doing the work manually. Yes, humans make mistakes and do shoddy work to, but humans screw up in predictable ways. The mistakes AI makes aren't ones that humans typically make. For example, people talk about AI use in my own field, law, as though it could eliminate 90% of our work. Aside from the fact that people who say this don't seem to have much of an idea about what lawyers actually do, the well-publicized boo-boos of lawyers submitting filings to the court that contain hallucinated cases that happen to perfectly fit with the case at hand aren't the kinds of mistakes lawyers typically make. In fact, if I fabricated a case in a motion because my argument was a loser or there wasn't anything on-point, I'd be sanctioned by the judge and fired from my job. Even if a first-year attorney submitted something to me for review, I'd assume from the outset that the cases he cites are real and that he's not misrepresenting them. Having to fact check every citation does not cut down on the amount of work involved. It may mitigate it somewhat, but that assumes that we're drafting from scratch; most motions I file are similar to ones that I've filed before, and I'm not going to reinvent the wheel every time I need to file one. AI drafting could theoretically give me scaffolding from which to work, but even then I can probably find a motion someone else has filed if I'm looking for a bare template. And that doesn't even get to the problem that I need to be prepared to argue the motion and am thus going to have to do the research equivalent anyway. So AI may shave some time off of drafting motions if I need to draft something I haven't done before, and even then it's not like I spend a ton of time drafting motions to begin with.
I know what you mean, I've run into this myself a few times, but I don't think all workflows are affected the same way. For images what you said harkens back to the old graphic designer's lament, when a customer really likes what you did, they just want a "small adjustment" like having a character face the camera, instead of facing sideways. With AI maybe the changes you want to apply aren't quite so drastic, but when the output is a flat unlayered image, making a small adjustment will indeed add work.
I don't think this applies to text output, though. Text is easy to modify, so if the AI gets off track you can correct it cheaply. Code comes with it's own set of traps, where it's possible to write something that makes sense when you read it, works, but contains so fatal flaw. Early on developers reported that whatever gains in code output are made up by the extra time spent on reviewing it. It's still an issue to some extent today, but in my experience nowadays AI just comes up with dumb ideas you never asked for, rather than doing what you wanted, with fatal flaws attached. The former is a lot easier to filter.
Either way, the proof is in the pudding, and I'm telling you reproducing the entire Adobe and Microsoft suite would have been a gargantuan undertaking as recently as last year, or even a few months ago. And it's not theoretical, we have actual examples of intelligent, driven people, with an ideological hate-on for big corporations taking their best shot at reproducing these suites to give them away for free. They've put countless man-years into it with laudable results, but since they don't have the same amount of resources they usually couldn't keep up with industry. Their work is now getting btfo'd by AI, which you can say is a far cry from btfo'ing Adobe itself, but I think it shows that all these arguments about how AI only seems impressive when you first use it, or that it makes errors humans wouldn't and this somehow costs more work than it saves, are all flawed.
I don't know about employment in particular either, people adapt, and it's not like the introduction of power tools caused massive unemployment in the construction industry, but we are definitely looking at a similarly sized disruption.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link