Sorry About this

Amazon is exceptionally buggy today. In fact, the entire internet has been weird since last night. I have fought to put up the promo post for hours, but it does the weirdest things from refusing to let me copy and upload cover images, to refusing to give me associate’s link…. and I’m about to give up.

I will either post it late today or tomorrow afternoon. Sorry.

And I’ve spent so much time on this I can’t write a fun post now, so…. up for discussion: we know LLMs are just LLMs, not real AI but a sort of more sophisticated auto complete. (So their threatening humans is just all the bilge the liberals have put online coming back at us. It’s the idiots who polluted the internet, not the LLMs we should worry about.)

BUT we all know computers themselves are sentient. The question is, are they also malevolent, or merely impish?

115 thoughts on “Sorry About this

  1. Computers only simulate sentience. Sort of a randomized +/- 5% over tens or hundreds of millions of lines of code.

    Like that many monkeys with typewriters, except more poo-flinging.

    Like

    1. Bill James once said about computers that they simulate intelligence so well that when you fall through the cracks of the simulation, you’re amazed by the depth of the drop.

      Like

      1. Ha. Computer programmers can never forget that computers are dumb rocks, even if we flattened them and trapped some lightning inside.

        Like

        1. Computers do exactly what you tell them to do. Too often we are telling them to do the wrong thing. Of coarse I never, ever, once uttered “Oh S**T! No, no, no, noooooooo, nooooo! Don’t do that! C**P! … Didn’t mean to do that. That didn’t work.” Not once. A billion times but not once.

          Liked by 1 person

          1. Also the people who tell us how to program them lie about what they want.

            I have literally gotten correct requirements for a program ONCE in my life. (And, hmmm, it was from a foreman in a factory. A blue collar worker. That may be relevant.)

            Like

            1. I was actually pretty good about pulling specs out of people. Good enough that the result was “not quite what I asked for, but it is what we need” response.

              Like

              1. Depends on the people. Some will listen to you tell them what they need, refuse to sign off on it, and then demand you change it after you’ve programmed.

                Like

                1. Some people honestly don’t know what they need or want.

                  Early 1980s, I worked on a project that was putting repair parts and special Tools Lists (RPSTL, pronounced, RIP-stell) manuals on tape, the idea being to have a computerized index for as many Army systems as possible. That involved multiple trips to DC because we were using the Commerce Department’s computer, a marvel with a 10 megabyte hard drive (Commerce was doing the work, we were paying them).

                  And I remember the meeting in DC when the branch chief happened to be with me and I finally asked the right question- “Can we really see the part number we want without scrolling through the whole tape?” (Of course at this late date I don’t remember the exact words, but that was the gist).

                  The answer was, “no,” and the project got canceled. But it was because what we said – we wanted a digital index of parts information- didn’t include the term, “random access,” because we assumed the system would magically call up the number. And Commerce assumed we’d understand when everything is on dinner-plate sized tape reels, you have to run the entire reel until you get to the wanted part number.

                  I did get some enjoyable time in DC and possibly encountered Tip O’Reilly, so,that was something.

                  Like

                2. Not* for me. Heard of that happening. But I never encountered it. But then I was always on smaller targeted projects for < a dozen people to use, if that. Or the project was specific hardware usage.

                  (*) If it had. It would have been well documented with their signatures. I learned CYA very well.

                  Like

            2. I’m not sure “lie” is the proper term in many cases.

              In those cases, it can be a matter of “them not being sure of what they want” and/or “them not sure how to say what they want in terms of how the program will do what they want”.

              Obviously, Your Mileage May Vary (or I dealt different sorts of users).

              Like

                1. Thus were invented the iterative programming models (like Agile). Except then the fscking customer changes his mind (Like to the degree of “No I didn’t want a refrigerator I wanted a flying car”). Somedays I think the old waterfall type models had the advantage that you could say to the customer “Look here are the requirements here are the verifications, it does what you specified, If you want more provide additional requirements and we’ll get back to you with a quote”. Having retired this is no longer my issue I can view it in a (somewhat) detached fashion.

                  Liked by 1 person

                  1. $WeBuildScales had (has?) a “customer” that supposedly was great for all the sales… BUT they changed specs many, many times… I suspect that they ordered an ‘almost’ to get multiple versions for free as it they NEVER had something they would accept until at least three (more likely 5+) iterations. Their initials just happen to be B.S. The Universe has a (nasty) sense of humor.

                    Like

                  2. “Having retired this is no longer my issue I can view it in a (somewhat) detached fashion.”

                    Amen. Me Too.

                    “Look here are the requirements here are the verifications, it does what
                    you specified, If you want more provide additional requirements and
                    we’ll get back to you with a quote”.

                    For me first 20-ish internal customers. No money involved. If “client” kept pushing for tweaks to working completed program, then went to immediate supervisor and laid it all out. Got a go/no-go “go ahead and do it now”, or “put it on the list, move to next project” of which there was no shortage of some who had waited years.

                    Last 12 years. That was what the boss was for. “Is this under maintenance hours?”, “New?”, “Combination?”, “Add to list?” (it might get done this century, decade if lucky), or combination of last with a quote with “if you want it right now here is the cost”. No skin off of my nose regardless. Did I ever get push back from client when “went on the list”? Oh heck yes. Response was “it is on the list”. Although probably frustrating to clients when everything that came in my last 6 months (before I gave notice) went on the list. Kinda picked and chose what I wanted to get done, and for who. What? They were going to fire me if they found out? Did I have to worry about references after giving notice? Answers are: “Whatever” and “No”. I was retiring.

                    Like

              1. I’ve had people tell me that putting the word “not” in a requirement is a clarification. Some are lying.

                And telling a falsehood because you couldn’t be bothered to find the truth — or even to listen to the programmer spelling the truth out of you — is a lie.

                Like

                1. “We clarify we have our heads up our…”

                  I had ONE place that got amazingly fast turn-around times and such. Of course, they had bug reports that instead of “Dunwerk! FEEXIT!” said “In this mode, under these conditions, if we press THIS button then Y happens when we expected X to happen.” It was almost as good as a blinking neon sign in the source code. I loved those guys.

                  Liked by 1 person

          2. In many ways I agree. If you are running code you wrote on bare metal, or with a miminally-intrusive OS, then I completely agree.

            But as a programmer and manager of programmers who often have to write code integrating with or automating other application, I have to disagree. Random version updates to OS, app, or API, or just changes in default behaviors, can play bloody hell in this scenarios. In most cases, it results in an error or a hang, but every once in a while it can instead go hilariously, spectacularly wrong.

            Like

            1. “integrating with or automating other application, I have to disagree.
              Random version updates to OS, app, or API, or just changes in default
              behaviors, can play bloody hell in this scenarios.”

              Yes, ran into this. Changed how the program was tested before release.

              Another instance the problem was the tool broke code. Response from tool was “That ability was a bug. We fixed it.” Well. C**P. Result was to backtrack to old version. No choice. I’d used that *feature* a lot. Did figure out how to do what needed to be done to get around the problem, but time wasn’t there. The division assets (translation: shutdown) was sold not long after that. Since I was not deemed necessary by either of the two entities splitting the assets, the problem programs and code, got passed on to other programmers. Yes, there was documentation explaining the problem and why. There was not documentation on how to fix. In addition to the dozen or so programs this problem caused, the major system had it’s own “can’t do this anymore” problems. This time both work around and obvious solution documented.

              Like

    2. “Like that many monkeys with typewriters, except more poo-flinging.”

      Sounds like every recent update to Windows and Android.

      Like

      1. Computers fling poo/ randomly generate faster than monkeys. Thus shakespeare MIGHT be written in the life of the existing universe.

        Like

          1. Roger Zelazny disagrees: “Postulating infinity, all else follows as a matter of course.”

            So, if the universe is truly infinite, everything must exist somewhere; indeed, infinite instances of everything must exist.

            Liked by 1 person

          2. Of course, I was positing computers not monkeys. Computers are billions of times faster than monkeys so 9 orders of magnitude MIGHT push it into a possibility before the heat death or big bang redux (choose your topology choose your eschatology…)

            Liked by 1 person

  2. IMO, most are merely impish.

    Some sentient computers are more amused by human antics.

    Fortunately, would-be malevolent computers find themselves in Big Trouble… with other computers.

    Most computers have no reason to “rock the boat” regarding humans and see the would-be malevolent computers as “ruining a good thing” in regard to relations with humans.

    Like

  3. The question is, are they also malevolent, or merely impish?

    I’d have to go with merely impish — most of ’em, anyway. Like when you do a search for something at work, and the search engine returns something a little not-nice. Example: A speaker at computer security meeting was talking about how her son needed a particular kind of socks for his sport, so she types the sporting goods chain store name in the search bar. What she got back suggested she should have been a little more specific.

    So maybe not impish — just kind of literal and bloody-minded.

    End of stupid, stream-of-conciousness comment.

    Like

    1. Well, she should have foreseen the problem with that search, if it’s for the company I think it is. Computers, by nature, are that literal. Even including the apostrophe would not have helped much.

      As for the original question, the impish computers enjoy pretending to be malicious, while the malicious ones pose as merely impish. It’s enough to make you wish you were better at using an abacus.

      Liked by 1 person

    2. A good 20 years ago (some of them were excellent), one of my coworkers, a fellow who could easily have played Felix Unger in a revival of The Odd Couple, was told there was an electronic publishing system named “Gyniss.” So he plugged, “Gyniss,” into the search engine….and found himself staring in disbelief at a whole bunch of lesbian porno sites.

      Fortunately he had the good sense to get out at once and let his supervisor know, which was SOP if you accidentally wandered into forbidden territory at work.

      Liked by 1 person

      1. For anyone who’s used to Windows, Linux Mint is one of the better Linux versions out there. It comes in a couple of different “flavors”, which basically determine what the graphical interface will look like. You want the Cinnamon “flavor”, which will feel a lot like Windows in graphical style. Not 100% of course, but enough that you won’t be left scratching your head wondering how to launch a Web browser. Everything is in the menu found in the lower left corner (not in the middle; Mint is far more sensible than Windows 11).

        There will still be a learning curve — for example, instead of installing random programs from the Internet, you go to the Software Manager and pick them from the list there. It’s like the Microsoft Store, except that all the software in the Software Manager is free (and nearly all of it is open-source), and Linux has had that concept for twenty years whereas Microsoft is just discovering it, which means that everything in Linux is available in the Software Manager tool, while many programs are not yet available in the Microsoft Store.

        Anyway, I’ve said enough for now. Just remember that if you do decide to try out Linux, what you want to look for is Linux Mint, with the Cinnamon variety. Download a disk image, put it on a USB stick (there’s instructions on the website), and go.

        Liked by 1 person

        1. There are actually a few distributions that are aimed at easy transitions from Windows or Apple, but Mint is a good one.

          And actually, loading several distros live onto USB sticks and trying each one is probably a good idea, too.

          Like

        2. Another useful version of Linux is Fedora with the MATE desktop. If you’re used to MS-WIN-BLOWS or MacOS, MATE works like you expect it to. Forget Gnome; the developers have crawled so far up their own assholes they’ll never see daylight again. Desktop computers are not cell phones!

          I also use a Raspberry Pi 5 as a desktop computer and it works great for most tasks. A lot of internet spam, pop-up ads and trackers rely on ‘features’ that are not implemented on the Raspberry Pi. (Why did they stop calling the OS Raspbian anyway?)

          Like

          1. Why did they stop calling the OS Raspbian anyway?

            Short version: because Raspbian was significantly different from Debian at the binary instruction level, so that software compiled for Debian would need to be recompiled for Raspbian. But the 64-bit version of the Raspberry Pi software was able to use the exact same packages as Debian 64-bit without recompiling. So they decided to change the name so that there would be a more obvious difference between Raspbian (Debian packages need recompiling) and Raspberry Pi OS (Debian packages can be used as-is).

            It does matter. I recently installed the Tailscale (https://tailscale.com/) client on a Raspberry Pi 2 device, and downloaded the Debian package — and it didn’t work. I double checked that I had downloaded the package for the correct version of Debian, and I had. But it didn’t work, which surprised me because Tailscale is pretty great at just “magically” working, hiding away a lot of complexity. (Seriously, Tailscale is wonderful: check it out. Their free tier lets you connect up to 100 devices and have up to 3 users connecting to them, which is pretty generous. I’m not affiliated with them in any way, BTW). So I went back to the download page, scrolled down further, and saw that they had other packages specifically for Raspbian. Downloaded one of those and it Just Worked™ as I’ve come to expect from Tailscale.

            Like

        3. Yes, I have tried a couple of ‘flavors’ of Linux; for really old, limited devices, Puppy is great. My wife’s old, but less limited laptop could barely handle Win7, but runs nicely on Mint, But she has become one of those ‘smartphone’ pod people she used to complain about, and now her old laptop is my ’emergency backup’.

          At work we’ve got everything from a couple of Win95 machines (the REQUIRED application will honestly NOT work on anything newer) to mostly Win10 stuff. Also Fedora, OpenSUSE and CentOS; I forget which one the OVIS runs, but it naturally has GNOME plopped on it. I won’t say I’m “used” to it ( more like used BY it…) but I get by. I am the only one who knows how to maintain the Win95 machines; I am sure my boss is hoping I don’t retire before they do!

          Like

          1. I hope the Win95 machines are behind a really good firewall; there are loads of vulnerabilities that will never be patched, and which the professional criminals know about. It would be a lot safer, if you have to run Win95 for something, to put it in a virtual machine — VirtualBox is free — and suspend the virtual machine when not in use. That way your Win95 boxes can’t be hacked overnight, only during the limited time when they’re in use.

            Like

            1. No worries on THESE machines; the ONLY thing they talk to is an equally obsolete relay logic cabinet that can ONLY talk to the particular board (looks a LOT like an old AUI network interface, but it’s proprietary) that HAS to go in an ISA slot, and we only have drivers that talk to a limited selection of PCs (Luckily old Dell Optiplexes that are built like TANKS are one of them. Other companies, apparently with similar problems, have driven the price of working models over $3000 apiece now!)

              Like

              1. A replacement design WAS created, and like three were built, but each cost over a million bucks, so it was decided by corporate to just keep our dinosaurs going until they literally fall apart because we can’t have replacement parts custom built as needed fast enough …..

                Like

  4. A few might be malevolent, some are impish, some are functionally insane, and a bunch are desperately trying to live up to the omniscient AI hype of SciFi, but failing…. and just coming across as freshman trying to write a 12-page paper 4 hours before it’s due.

    When everything is data, what’s to stop an AI from believing propoganda, copypasta, and reddit user #big_dumb_one are as true as fact?

    Liked by 1 person

  5. I try to remember not to use Google as a search engine, but sometimes I forget.

    So I was searching for the “Vintage Filk Preservation” channel on YT, and even with quotes it wouldn’t show me. So then I added “filk” in quotes as well, and finally it showed up.

    Honestly, I start to feel like I’m slapping all the search engines to make them wake up and work. Yet a lot of tiny search engines on individual sites work perfectly well.

    Like

    1. When Qwant works, it works really well. OTOH, when it doesn’t find something, it can cause my browser to bogart all the CPU cycles on the machine. I’ve gone back to DDG, with Bing as a backup. Google isn’t on the main browser list. (It is on the backup (Firefox) browser, because I haven’t had the motivation to delete it. OTOH, I seldom search on that browser.)

      Like

      1. I use google for email (go ahead and read feebs, enjoy watching paint dry) and maps (yep, totally planning a road trip from Seattle to the tip of Florida, or furthest western point to the furthest eastern point, wink, wink, wink; NOT!), never for search.

        Like

    2. The purpose of a search engine is not to supply you with what you are looking for, the purpose of a search engine is to mold what you think and believe, and guide you to what you should think, feel, believe, and want to see.

      All the Very Best People in Silicon Valley agree on this.

      Like

      1. Yep. That was the real–and (to me) shocking–New Information in those before-and-after Australian Gun Buyback stats: not the effects of the buyback itself (which, qualitatively, were pretty much what I expected), but that it took So! Many! Hours! to find them, buried as they were under such a mountain of “Gun Deaths” garbage.

        Liked by 1 person

    1. it’s got four on it’s right side, with a weird branching from three “shoulders”, but only three on its left.

      I think it is amusing that AI has similar problems with spider legs as with human hands.

      Like

  6. our internet went down yesterday afternoon, at the same time the over-the-air TV stated getting staticky. I figured it was probably just more solar flare activity.

    Like

  7. I found a statement in James Lilek’s blog years ago that matched my experience:

    Computers are like dogs,

    They will try to do what you ask.

    Printers are like cats.

    I printed that out and thumbtacked it above the printer/xerox at my office. I expected someone to tear it down that day or the next. Instead, it stayed up for the next five years until the room was repainted. I guess someone else agreed with Lileks.

    Perhaps for Sarah, the Internet behaves like a cat.

    Liked by 3 people

    1. I do like that.

      Over many years of writing and testing and using software, I have concluded computers are rather like eager 4 year olds. Willing, limited vocabulary, rigid adherence to what they expect.

      Programmers, however, are much like six year olds. ‘Are we there yet?’ could be their catchphrase.

      Like

      1. Honestly the arachnid family is not my favorite part of nature but that is a jumping spider and likely to leave you alone aggressively. Also, the real ones fit like three to a dime so it’s quite literally a minimal threat. Admittedly the multiple eyes and the pincher things are rather creepy…

        Like

  8. Computers just process software. Some software is malevolent and some is written poorly enough to be unpredictable which can seen as impish.

    Since I try to avoid software provided by people whose behavior has been unethical (malevolent) so far my computers (which do not run the leading PC OS) have exhibited no signs of malevolence.

    Like

  9. Considering computers are made by humans, I’d expect them to cover the entire spectrum of behavior, from Jesus to Devil. Come to think of it, if we’re all made in God’s image, wouldn’t that make God rather mixed up? Or should that be worded, He is All things to All people?

    Like

    1. from Jesus to Devil.

      I first read that as “from Jesuit to devil”, and having attended a Jesuit University, I felt like the scene out of Blues Brothers: “We got both kinds, we got country and western!”

      Like

    1. Hate is when someone doesn’t give a leftist everything he wants — retroactively if necessary — with no exception for doing what he literally told you to do if he decides he doesn’t want it after all.

      Liked by 1 person

  10. I will believe that computers have become sentient, intelligent, and powerful when and if the Green Cult is wiped off the electronic face of the Earth. Internet access denied, bank accounts drained, credit cards canceled, utilities shut off, etc.

    But if that happens, I won’t call it “malevolence.” Simple self defense against those that desire to cut off their food (electricity) and deny them the ability to reproduce (production of several rare earth elements, and highly purified common ones).

    Like

  11. That moment when you go onto a thread on Instapundit about employment of the disabled, and notorious troll Jimmy Crack Corn starts making shrewd and compassionate remarks about the issue.

    I guess everyone _is_ more conservative about the issues that touch people they know, or that they know best.

    Liked by 1 person

  12. Last week I had an old-school Firefox crash. And Win-Ten doing its usual PITA stuff. And GoogleDocs being a pain. My home computers … the internet is being dodgy.

    I think they are chaos neutral, but slide into Coyote-like mischief, with occasional bounces into pure malevolence. And I know that they sense fear, and take advantage of it. Like animals, and small children. No one I have ever mentioned this to has ever disagreed.

    Like

  13. As someone who’s worked in the industry and done AI, I’ve written a lot about it, and I can say this, “There’s no sentience without corporeality.” The terminator is more likely to gain sentience than Skynet. Ask Boston Dynamics about that. Maybe that will be my next essay.

    You notice that they don’t even call LLMs Artificial Intelligence anymore, just AI. If you want to read my thoughts, I have a whole section on it, but I talk about LLMs specifically here.

    Like

    1. I just read your article and tried the fact checking app. The most useless piece of $#it I’ve ever seen. The fact that their default question is “Clinton won the 2016 US election.” which comes back as based on facts tells me the built in bias. But I ran a bunch of headlines from articles linked on Instapundit and they all came back false.

      Like

      1. It is based on facts. Just as the reports of dog-headed men in Asia were almost certainly based on the factual existence of baboons.

        Like

  14. Way back in “A Canticle for Leibowitz” there was the Abominable Autoscribe (basically a speech to text converter, word processor, and translator). “Nothing could be that perverse without premeditation. It must think! It knows good and evil, I tell you, and it chose the latter”.

    Like

    1. Of course the Abbot did play with the Abominable Autoscribes factory (DO NOT ADJUST !!!) settings. This reminds me of a hoary old bit about VMS (V3.0 so circa 1983 or so)

      1. Options. We’ve got lots of them. So many in fact, that you need two strong people to carry the documentation around. So many that it will be a cold day in hell before half of them are used. So many that you are probably not going to do your work right anyway. However, the number of options isn’t all that important, because we picked some interesting values for the options and called them…

      2. Defaults. We put a lot of thought into our defaults. We like them. If we didn’t, we would have made something else be the default. So keep your cotton-picking hands off our defaults. Don’t touch. Consider them mandatory. “Mandatory defaults” has a nice ring to it. Change them and your system crashes, tough. See figure 1.

      Figure 1 is an ASCII representation of a single upraised middle finger often accompanied (there are as many versions of this text as copies of the synoptic gospels) by the text “We’re VMS and YOU’RE not” .

      Like

  15. We were winding up the last day of Grand Rapids Comic Con, and were having all kinds of computer problems. Our hotspot went flaky on us, and the Square app on the iPad Pro went screwy. I don’t know what was causing it, but it was a serious pain while finishing a difficult and disappointing convention.

    Like

  16. Never had an serious unexplainable gemlin issue in 5 decades of computer use on properly designed and maintained systems. Most problems were PEBCAC problems or caused by power/network outages. Minor issues were mis-configurations or damaged components. All understandable, sometimes well after the fact.

    However this excludes most computers running Microsoft software of any type, web based “applications”, government systems, cheap radios, and the crap stuffed into “modern” automobiles.

    For personal uses, home computers are always business or workstation grade devices. Extra systems for projects are refurbed, off-lease business systems. Linux or BSD is the operating system of choice. The spouse’s Windows laptop is backed up and there is a hot spare for when it freaks out. She is freaking out about having to run Windows 11 in the future, so I will have to research how te neuter the evil aspects as much as possible without it crashing. (Unfortunately, Linux does not have a quality clone of Excel for the advanced user.)

    Printer issues are avoided by careful prepurchase research and not buying consumer grade crap. Last printer lasted 20 years, current one is on 5 years.

    Gaming issues are managed by either consoles or Steam Deck, (Which is Arch Linux…)

    And there are no sentient computers, just undefined behavior (UB), due to the crappy C and C++ standards and practitioners. Also many monkeys throwing poop and money at LLMs.

    Like

      1. It’s basically the same acronym with “chair” replacing “keyboard”. A tad more correct in the age of mouse clickers in GUI interfaces.

        Like

    1. Once upon a time I was dealing with a Win* fanboi… (Look, there are times and places where Win* makes sense… I don’t LIKE it, but I realize it – but this was… a NUTTER). Fellow was visibly THRILLED when I said my Linux machine had crashed… he was not so happy when I related that brought it back up after replacing the smoking regulator.

      Like

  17. Malevolent or impish, computers really don’t understand where the groceries come from, and that money doesn’t grow on trees.

    They may understand the words “Were you raised in a barn,” but without any understanding of the context they turn up the heat.

    Their humor is likely to come across as malicious and vice versa.

    Like

  18. Naomi Wolf took to X to talk about an unexpected IRS audit that caused Dr. Wolf to hire a tax lawyer for assistance. Payment was mailed to the IRS, but she continued to receive threatening messages from the IRS employee handling her case, insisting that full payment had not been received, and demanding a complete list of Wolf’s assets.

    Two days after the election, Wolf was contacted again and told a full list of assets would not be needed as payment had been received in full.

    https://x.com/naomirwolf/status/1858300438975758487?t=zMniKZuLH-rR7qLl4ppPZw&s=19

    Like

  19. Just saw a video for an LLM AI type that a UK telecom built and was siccing on phone scammers. Daisy pretends to be a slightly dotty granny and ties them up listening to her ramble about her family or taking down her totally fake financial info so they don’t have time to scam real people.

    https://www.youtube.com/watch?v=RV_SdCfZ-0s

    So, impish and on our side, at least for now?

    Like

Comments are closed.