Today's post is going to be a bit of a mixed bag, I think.
Not a Grab-Bag. I have a sort of format for those and this isn't going to fit it. It's just a few things I wanted to post about that probably won't make full posts of their own.
Then again, maybe one will blow up into something bigger as I write, in which case I'll just come back and delete this introduction and no-one will ever know! <Twirls mustachio. Supervillain laughter.>
Always On
First, something that definitely isn't worth a whole post. I just want to moan about it. Unlike some people, Nimgimli for one, I've had absolutely no technical problems with Nevernesss To Everness so far. No bugs, no UI glitches, no performance issues. For me, playing on PC, it runs as smoothly as any viscous liquid you care to name.
Playing is no problem. The problem comes when I stop. In the last few days - I bet since one of the frequent updates, although I couldn't nail down exactly which - whenever I log out of NTE, about half a minute or so later Windows tells me it's "run into a problem" and needs to reboot. That would be annoying enough but it turns out Windows can't reboot and I end up staring at a black screen until I switch the power off and restart, after which everything works perfectly until the next time I stop playing NTE.
Apart from being annoying, I worry all this sudden stopping and starting will damage something, so I googled for explanations and fixes. First, I did it the old-fashioned way. I checked reddit threads and watched YouTube videos but no-one seemed to have the exact problem I did and nothing they suggested seemed particularly helpful, so I thought I might as well let Gemini have a go, since it kept on offering.
Gemini was extremely co-operative. It asked pertinent questions, gave me lucid explanations, offered fixes, walked me through what to do when I had difficulties implementing them and basically acted like the best kind of IT department I've ever had to speak to (And I've spoken to plenty.)
All of which would be great if the solutions Gemini provided had worked. They did not. Oh, they worked in the sense that all the commands and instructions were accepted when I followed them and they did what they were supposed to do. It just didn't stop NTE crashing my PC on exit.
But then, neither did any of the non-AI fixes and suggestions I tried. If it was a football match it'd be a no-score draw. (But then, I just used Gemini to fix a perpetually annoying issue I have with Blogger getting the color of links wrong and it sorted it out perfectly in ten seconds, so I guess AI wins in injury time.)
Of all the various possible reasons offered, by far the most likely seems to be a conflict with the Anti-Cheat software NTE uses. From long experience with online games, the most likely fix is going to be putting up with it until the developers patch again and it magically goes away. Until then, I might just try shutting the PC down immediately I log out to see if I can beat the crash. That'll be fun. [Edit: Tried it and it works so that'll be my temporary solution for now.]
Had Gemini's fix actually worked, I might have been here today singing AI's praises. That'd be a popular post, I'm sure. If anything, anti-AI sentiment seems to be growing. It used to be mostly in my gaming and music feeds but now it's increasingly present in just about anything I read. As for positive sentiments regarding our would-be artificial overlords (That's Google and Amazon and whatever Elon Musk is calling himself today rather than the inert and blameless software itself, of course.), those seem to be very thin on the ground indeed.
Search Me
All of which does make me wonder, even more than usual, how this is all going to pan out. I heard the rumor that Google plans to replace search entirely with some kind of Agentic AI (I do love that word - Agentic - don't you? Doesn't it just ooze futurity? Algorithms never had that kind of PR.). It sounded a bit worrying so I checked (Using Google Search, inevitably.) and it turns out to be the usual kind of hyperbolic over-exaggeration humans have been using to get Eyeballs or Clicks or whatever the metric is these days since at least the day Buzzfeed went live. Which was exactly twenty years ago. I just checked. (Google>Wikipedia.)
In fact, Google Search continues as before, according to a statement Google gave USA Today, who bothered to ask them, but there will be a new All-AI front end as well. That, inevitably, will be Google's new focus and I'm sure it will be the first/main thing you see, which means most people will use it without thinking any more about it. I imagine their hope is that Search itself will wither away from neglect and disuse and they'll be able to discontinue it at some future date when no-one cares any more.
Will that happen? Hard to say. How did Google take over from all those other search engines - AltaVista, Netscape, Yahoo and the rest - in the first place? It was faster, more accurate and more comprehensive, that's how. People used it, found it did the job better and stopped using the older search engines.
Have people changed that much in a couple of decades? If they find the new AI Agents are worse than the search they had before, will they not move away from Google to something that gives them what they want? Isn't it just handing a huge opportunity to a new "Traditional Search" provider to come into the market?
Or, much more likely, will most general internet users find AI means much less fiddling about and reading websites and a lot more getting quick answers that work well enough, often enough, which will be plenty to keep almost everyone at least happy to go along with it? Too much effort just to get back to something they probably won't miss anyway.
So, yes, I imagine AI Agents are going to replace search if only because I'd bet the huge majority of users never really liked searching to begin with. It was always a necessary inconvenience for most people and I'll bet they'll be glad to see the back of it. People who actively enjoy searching as we've known it have to be a pretty small minority of web users, surely?
I'm kind of on the fence about the whole thing. I definitely don't hate AI. I just wish it was better. Maybe it will be, one day. Or maybe the current technology, which seems to be part brute force and part black magic, is a dead end and it'll never be entirely reliable. I suspect that's more likely but it's too soon to jump one way or the other.
You Want Me To Draw You A Picture?
All of which brings me to a little discussion that took place in the comments on a post over at The Friendly Necromancer, where Stingite was talking about feeling guilty for using AI art to illustrate his (Other.) blog, rather than, for example, hiring an actual artist to do it.
I said in the comments that it's a notional argument. No hobby blogger is ever going to commission an artist to provide illustrations for posts except on an absolutely exceptional basis. I must have read tens of thousands of blog posts now and I can't remember ever seeing it done. It didn't happen before AI so AI isn't stopping it happening now. No artist is starving because a blogger stopped commissioning spot illustrations for their posts.
Very, very occasionally I have seen someone commission a piece of art to be a permanent feature on a blog. I remember Belghast doing it for a masthead a couple of times and I have a vague idea one or two others may have done something similar. But no-one who posts several times a week is going to pay a commercial rate to a professional artist for even one illustration per post, let alone the half-dozen or more most people who use pictures at all like to throw in
And that in turn got me thinking about The Olden Days. I'm not talking about Ye Olde Webbe of Yore that so many people, most of them barely old enough to have experienced it the first time around, seem so struck on bringing back. I'm talking the way things were before the worldwide web even existed.
When I came back from college in the early 1980s, one of the first things I did was start a comics fanzine with my then-wife, a friend of ours and the guy who owned the comic shop I worked in. We put out seven issues over two years and then our friend took over the editing and publishing of a bigger, more successful 'zine, which he eventually turned into a semi-pro operation. I switched to writing for that and we pulled the plug on our own zine.
Every issue of our original zine was stuffed with what we called "Spot Illos" - either decontextualized images, used to break up the text, or more targeted images, intended to support it. We also had comic strips sometimes and full-size cover art for every issue.
A minority of the pictures were drawn by my wife, who was a great comic artist and should have made a career out of it, but most were done by people who read our zine and who were active in comics fandom at the time. Some of them already had a foot in the door of professional comics publishing, some went on to be professional comic artists later, but most remained hobbyists and amateurs.
Whatever their status and ability, no-one got paid a penny. No-one expected to be paid. Paying people for art that wasn't going to be sold for a profit was not a thing anyone did, wanted to do or even thought about doing. All people wanted was to see their work and their names in print. If they did have professional aspirations, they'd add it to their portfolio so they could at least show potential employers something they'd had published but most of our contributors weren't even that ambitious. They just liked to draw and enjoyed sharing the results.
If something similar was part of blogging culture, the way it was always part of the 'zine culture I grew up with, no-one would need AI to draw them a picture. There'd be no shortage of people happy to provide it for free. We always had far more submissions than we could use.
And we had a smaller readership than many hobby blogs, too. From memory, I think our print run was about 300 at the peak although the semi-pro zine my friend ended up editing and publishing ran to ten times that eventually. And I don't believe he ever paid anyone anything, either, until a bigger publisher picked him up and gave him a budget to go pro with an actual comic.There could be a place on the web where bloggers could ask for images to illustrate posts and artists could supply them for nothing more than credit and a link. The technology has been in place for years to allow something like that to grow into a global free exchange of talent. Granted it would never be quite as instant and frictionless as generating an AI image but the results would be so much better it would be worth the wait. Probably. Although now I think about some of the pictures we published, let alone the ones we didn't...
Maybe something like that does exist already. I know it does for paid, commissioned art. If it does, though, the evidence has never shown up in any blog I ever read. And I'm certainly not offering to set up any such kind of website myself, although ironically I imagine I could get an AI to to code it for me if I was. They're supposed to be good at that sort of thing.
And even if someone else did all the donkey-work, it wouldn't be great for me as a user anyway. It would better suit people who write their posts with at least a little lead-time. I tend to bash mine out on the day and I don't think there are many artists out there who'd be happy to get a request after lunch asking them to knock out half a dozen pictures before tea.
That's how AI wins, I guess. It may be soulless but it sure is fast and it never complains or makes excuses. It never says "Do it yourself. I'm busy." Or fobs you off with "I've just got to walk the dog and do a bit of shopping. But I'll get to it as soon as I can. Promise!"
And yet, I don't use a lot of AI art here any more. It's not even because readers don't like it. When I do drop a few AI illos into a post, most people just ignore them, I think, assuming they even notice. As Stingite says, AI's much better at doing art than it used to be so it doesn't stand out the way it did.
No, it's more that I find it a bit dull, now the novelty value isn't there any more. I'll use it if I need to but it's purely functional, not the crazy thrill-ride it could be a few years back. I get better results dicking around with images in Paint.net, anyway, and that feels a lot more creative than writing prompts. I'm not at all sure it is but it feels that way.
Hmm. I seem to be wandering away from whatever point I had. Not that I expected the post to go anywhere but at least I got a few things off my chest. I had a couple more somewhat-related topics to talk about, too, but since this has clearly gone on long enough already (More than, probably...) I'll save those for another time.
Now... shall I use AI to illustrate this post? Would that be ironic? Post-modern? Provocative?
Or just plain lazy?




It's the 2nd point you make that would prevent me from ever paying someone to create images for me. I can barely wait long enough to proofread my posts before hitting publish, let alone wait a few days for an artist to create something for me.
ReplyDeleteIf I DID have to pay for images I just wouldn't have images. No one is paying me to write my blog, and I already spend money on hosting every month, so spending more on images would just make it more likely that I'd shut the whole thing down.
I also find it vaguely hypocritical that people say AI is stealing art when many of these same people are doing things like creating illustrations of characters from their favorite game or anime or whatever. Isn't that also 'stealing' someone else's design? This doesn't apply to folks who do original characters/art of course. But copying is how humans learn; it doesn't seem so crazy that the AI we created do the same.
But I don't understand all these legalities anyway. If I stand on a corner and do Wet Leg cover tunes and make $1000 in tips, do I owe Wet Leg royalties? (I'm using Wet Leg as an example cuz I learned about them from you.)
There's actually a lot of discussion about whether LLM will every get to AGI and smart people are looking at different ways to approach things. On the other hand I saw a fascinating story about how researchers created an AI to play Pokemon (not sure how Pokemon became the benchmark IP for this stuff) that could teach itself while playing. So it started out really poorly, but it improved itself over time via basically trial and error. Then when it finished the first game they set it on a sequel with its 'intellect' intact and it did much better there, and on and on.
So basically it was self-evolving, just limited in the Pokemon sphere of influence. But that was kind of a breakthrough since tuning models I guess costs a lot and this one was just tuning itself.
I didn't know Google had a max comment length so that's useful information. Surprised I haven't hit it before now.
DeleteI'm pretty sure very few people, myself very much included, could tell AI art from human art in a blind viewing. There's a whole lot of psychology on this sort of thing, though, as it relates to other things like food and drink and what matters most is expectation. People experience what they expect to experience so if they know it's something they don't think they're going to like, they won't like it. As long as there's the current demand for AI art/music/video to be clearly labeled as such, it probably won't matter how close it comes to non-AI equivalents.
The other thing, though, is that i already have a fairly strong antipathy for much human-created art, especially in the areas I tend to see AI operating. I've had decades of trying to avoid much of the kind of art that people using AI seem to love to get it to produce. It really doesn't matter much to me if it's human or AI if it's in a style i already want nothing to do with! You might have noticed that when I use AI art here and give the prompts at the end, they often include things like "1950s magazine" or "1970s comic book" as style guides. There's not a great deal of line art from the 80s on that I really like. I often compare the current AI hate to what people said about airbrush art in the 80s but I don't clarify that one of the people saying that back then was me! I didn't like it then and i don't like the AI version now.
On NTE, I'd love to play on Steam but i don't want to have to start from scratch again. Ditto Google play. If they allow existing accounts to transfer or cross-play though, that'd be a different story. I'm hopeful it'll patch itself right soon enough, anyway. That's what usually happens. And that puppet quest is very good. I have a post brewing that includes something about it.
I just checked and NTE does have full cross-play across all platforms so long as you use a PWE account so I might look into that a bit. Thanks for the idea!
DeleteThis comment has been removed by a blog administrator.
ReplyDeleteSo tempted to leave this obviously AI-generated comment in place for the sake of irony but I suppose I shouldn't.
DeleteThe issue with AI, as I see it, is that it's going to screw us all regardless. The peaple building it are happilly promising investors that withing a few years AI will be able to do "all jobs with economic value" better than a human can. If they are correct, that also means the enourmous buildout of infrastructure for AI will continue indefinitely, almost reversing recent reductions in CO2 emmisions and doubling or tripling the power bills of anyone that lets one of these new data centers get built in the same energy market they buy power from. We will be paying largely for the infrastructure buildout, but also likely paying higher prices per kilowatt hour regardless. For example, in some markets here in the US that have been overly welcoming (i.e., peaple in state government have been bribed), ultility bills that exceed monthly mortgage payments are becoming common.
ReplyDeleteBut let's say that they are wrong, and all this hype is completely unwarrented. That would mean that we are on the verge of seeing the most catastrophic stock market collapse since the one at the end of the radio bubble that triggered the great depression.
So yeah, I'm not surprised folks are pretty negative!
Supposedly Gemini is the most accurate of the major AI platforms. Gemini is the only one I have used much, since I am already hip deep in Google's ecosystem. I have to say it's pretty damn impressive. I have even used it for a few work tasks, and it did a great job with the few things I asked it for. I have also caught it lying to my face a couple of times, though it did at least apologize profusely when called on it . . .
If there's any truth to the claim that AI will "do "all jobs with economic value" better than a human" then the cost of electricity is going to be irrelevant because the entire property-owning and renting class will be out of work. That's always been the part I don't get - if the project (And it is a massive social engineering project at its core and was always meant to be from what I've read.) is successful, what's supposed to be the worthwhile outcome? Everyone focuses on the supposedly sociopathic billionaires driving the AI revolution but most of the people who got us to where we are had (And have.) grandiose ideas about changing humanity for the better. I just don't see how replacing every paying job in the world gets us to a better place. It's no wonder conspiracies involving the euthanizing of 99% of humanity take hold. It's kind of hard to see what the point of leaving anyone alive is going to be.
DeleteAlso, what happened to a) quantum computing and b) cold fusion? Either of those would at least make operating AI less impactful although it wouldn't keep anyone in a job.
A few quick points (I'll try not to rant too much):
DeleteFirst, I think a lot of the "anti-AI" mood is more against the current unregulated systems. Lots of techbros racing to be the big winner in a venture capital competition, without regards to the side effects or losses. If things crash hard, they will just go on to the next big disruption without much personal cost. If the rush was slowed down by say a factor of 4-10 times (still a LOT of investment) with some effective government oversight there might be less resistance.
As for "doing all jobs", I think that our societies should do a lot of rethinking about what the goals of governments. Right now they are heading towards techno-feudalism with governments mainly just enforcing contracts. We made a lot of social progress in the past couple centuries, but a lot of that can be revoked (or more positively, improved).
On AI "lying"--the AI has no sense of truth or falsity of its outputs. Essentially, it is telling a plausible story every time. One useful thing: try accusing it of giving a wrong answer *even when you know it is right*. Generally you will get an apology and a different answer.
Search engines: Ironically search is one of the few things that current LLM tech *could* help with a lot. The trick is to ask for sources or links rather than trusting the AI story to be a true one. LLMs are great at lumping together different terms for the same ideas: for example "type 2 diabetes", "hyperglycemia", and "high blood sugar" which conventional search won't do.
Finally, cold fusion is real, but there is no reasonable way to produce significant energy from it. Quantum computing is also real, but we are basically in the VERY early days quite similar to the vacuum tube era of computing. My guess is another 10-20 years before there is a commercial use for it.
All excellent points with which I agree. I had a news item about AI apologies bookmarked for ages, meaning to write a post about it, but I never did. most of them seem to have been toned down now anyway, so they don't chat away like your best and most obsequious buddy any more - thank god. I hated that. I did rant at Gemini yesterday, though, because of the annoying way the "Enter" key works there - basically it submits the text instead of acting as a carriage return - and Gemini agreed with me and sympathized, which is pretty rich considering it was the one causing the problem in the first place! If it really was sorry it would use its coding abilities to re-write the software and stop it happening - but then to do that it would also have to be actually intelligent...
DeleteThis comment has been removed by the author.
DeleteA few final follow up points:
DeleteI completely agree that the basic issue is unregulated AI development. At least some of the anti-AI sentiment is due to more and more peaple starting to realize that we are on the verge of letting five or six unelected tech bros become our feudal overlords in all but name.
I think you really hit the nail on the head. If our governments were working to navigate a path that would be of the most benefit to everyone, I would be a lot less pessimistic. However, here in the US our elected "representatives" seem to be in no danger of doing anything of the sort.
For the most part, they are pushing policies that benefit tech bros over the rest of us becuase they make large campaign donations. There are also consistant rumors (some even susbstantiated and getting mostly ingored by the mainstream press for some reason) of politicians getting wealthy (er) off of insider trading related to this and other policy matters. Not sure if things are going any better in Europe.
I also agree that if the technology were regulated sanely, it could be of great benefit, and that search is one area where it can be really useful. For example, Gemini is really good at helping you locate primary sources relevant to a topic you aren't already intimately familiar with. You of course have to read the sources it finds, since it will very confidently misrepresent them on occaision. However, I do find it to be an improvement over unassisted search engines when I have a really technical question. I can do deep research on a "new to me" topic in something like a quarter of the time it would have taken me two years ago.
I also am well aware that AI neither knows nor cares whether anything it tells you is correct. As everyone in this thread is likely aware, it's really a very advanced statistical model that predicts the most likely correct response based on patterns in text that has been fed into it (tokenized and used to paraterize a model). Becuase the model has trillions of parameters (for example, GPT 5 has about ten trillion), it produces a very convincing illusion of thinking. I know full well it does nothing of the sort.
However, when discussing the output, it's a lot easier to discuss a colloquialsim such as "lying to my face" than to describe the process accurately. Much as when discussing interactions with NPCs in a game, I don't talk about our conversations as the output of programmed routines . . . .
EDit" almost certainly" ...not "almost"
ReplyDeleteTo clarify, I did a bit research on overall trends in greenhouse gas emmisions recently. To summarize, due to accelerating adoption of renewables, annual CO2 emisions have almost stopped increasing. For example, in 2024 we barely emitted more than we did in 2023 (which is better news than it sounds like!), and if the trend were allowed to continue, we would see at first a levelling off and then a slow reduction of emissions in the next few decades.
Thank you for all the Beryl pics!
ReplyDeleteAnd yes, I do kind of wish that AI would tell a user to occasionally just "piss off and do it yourself". I know I'd laugh my ass off if it happened to me.
If I'm ever stuck for something to illustrate a post, Beryl's my go-to!
Delete