@lynessence@fastfinge And I'm gonna be one of those multiple people in a thread people because yes, that game was freaking nuts. If my sleep schedule becomes a casualty to baseball... I will absolutely enjoy the hell out of it.
@quanin@lynessence Tonight is gonna be interesting, with them all exhausted. And I'm happy we've managed to turn this thread into exactly what I was talking about. You know, as an example for anyone who hasn't noticed the phenomenon.
@fastfinge@quanin@lynessence Have you considered firing him and putting AI in charge? That seems to be the hot new thing that definitely will not go wrong.
@dhamlinmusic@lynessence@quanin As a large language model, talking about fire is irresponsible and against my ethical guidelines, as it may burn people. Perhaps we could talk about something else instead, like pink fluffy unicorns dancing on rainbows.
@quanin@fastfinge@dhamlinmusic@lynessence At least, from Ice Crhimp, you could go to a version with less bells and whistles, called frozen shrimp. And then, a very basic one called, breaded shrimp.
@mcourcel@quanin@fastfinge@dhamlinmusic Oh you mean while you're writing all kinds of AI music? Lol but seriously, I agree. I'm worried about the long-term effects on both the planet and our children.
@lynessence@dhamlinmusic@quanin@mcourcel I'm kind of not? Like, they didn't need AI to elect a fascist government to our south. AI might make things slightly better, or slightly worse...but mostly it'll just make things different. We didn't need AI to lose trust in the news media. Covid and vaccine conspiracies were alive and well before GPT. Bitcoin and cars were already destroying the environment. Cheap kids crap on YouTube was already destroying brains. AI is either a cure or a null factor.
@fastfinge@lynessence@dhamlinmusic@mcourcel I mean, all of those things are bad, absolutely. But this drive to lean on AI like people want us to is making us lazy in ways YouTube can only dream of. People at least used to google shit 5 years ago. Now they just ask GPT and GPT googles shit.
@quanin@dhamlinmusic@lynessence@mcourcel Nah, five years ago they used to ask on mailing lists for someone else to Google Shit. The lazy people were always, and will always be, lazy.
@fastfinge@dhamlinmusic@lynessence@mcourcel Even those people are now asking AI to google shit. A thing I hear regularly is "it will take me 20 minutes to find the thing. AI can do it in 2."
@quanin@dhamlinmusic@lynessence@mcourcel And they are largely correct, mostly because Google has become utter shit. Kagi is better. But half the time I find myself using AI because I don't want to sift through the advertising filled news website and download 18 megs of JavaScript and skip four autoplaying videos just to get a single fact. So I use AI because, in the world that exists, it's much easier than the alternative. And it's not that much less accurate, when judged against the accuracy of most of the garbage websites out there.
@fastfinge@dhamlinmusic@lynessence@mcourcel Eh. I can still make Google do my tellings. In fact, on the occasion I ask AI to look something up I'm fact-checking it anyway because I want my own deep dive.
@fastfinge@quanin@dhamlinmusic@mcourcel This is truth. I work in customer service, I have people calling me all the time to fill out Google forms that are perfectly accessible. But then they tell me "but I'm blind and this is inaccessible", and I secretly laugh at them in my head because they don't know that I am also blind and in filling the form out for them.
@lynessence@fastfinge@dhamlinmusic@mcourcel ROFL. When I worked for Dell I had someone try that with me once. The 30 seconds of dead silence when I told them so am I was priceless.
@quanin@fastfinge@dhamlinmusic@mcourcel I really really really want to say that one of these days. I just don't know if I would get in trouble or not. Probably not. But I can't afford to be jobless.
@lynessence@fastfinge@dhamlinmusic@mcourcel I highly doubt you'd lose your job. You work for a company that assists people. What you'd be doing is basically gold standard confirming that yes, this thing is accessible. You might offend the odd blind person who thinks "accessible" means "I know how to use it", but that's not a firable offense.
@quanin@fastfinge@dhamlinmusic@mcourcel Agreed. I don't think I would get fired. But I already have Aira explorers who hang up the phone when I answer it and refuse to speak to me, so best not to rock the boat. I don't understand what the issue is, I'm a nice polite Canadian. Maybe they are Republicans who don't like Canadians. Lol
@lynessence@fastfinge@quanin@mcourcel See I have run across this more than a few times, something should be accessible, but the entity that created it only tested it on something like JAWS with Firefox, or Voiceover with Safari, and it falls apart when you show up using Talkback on Edge, or Narrator on IE11.
@pixelate@fastfinge@lynessence@quanin@mcourcel I am trying to figure out why Facebook will not log in on Edge with #NVDA, the information is correct, works on Edge on my phone, even saved it there to have it auto fill, my best guess is uBlock.
@fastfinge@lynessence@quanin@mcourcel Somehow Talkback on Chrome either works flawlessly, or they manage to break it in ways even Google cannot figure out. I had Google Forms that someone managed to break for use with Talkback.
@dhamlinmusic@lynessence@quanin@mcourcel I think that says more about how fragile and broken TalkBack is. I have an Android phone for work and I had to replace talkback to get anything done. And it's still not as smooth as IOS.
@dhamlinmusic@fastfinge@quanin@mcourcel These are folks who are using windows with jaws or NVDA, and I know Google forms are accessible with those screen readers. I am using a Mac, which is usually more troublesome and less accessible.
@pixelate@dhamlinmusic@fastfinge@quanin@mcourcel I have heard that, but I've never run into any issues myself with DocuSign. But I suppose there are probably multiple ways to format a document and some are probably inaccessible.
@lynessence@dhamlinmusic@fastfinge@quanin@mcourcel I seem to remember one where the form fields weren't labeled, so it wasn't clear what it wanted me to enter. But then I could be remembering one the many PDF's I've had to look through at work. I absolutely hate PDF's.
@pixelate@dhamlinmusic@lynessence@quanin@mcourcel But making accessible PDFs is possible! All you need is fifty thousand hours of training directly from Adobe, two software packages nobody has ever heard of before, and for making accessible pdfs to be your full time job. Either that or you could just hire one of the people from the thousands of ads I get in my Linkedin messages every day from companies nobody has ever heard of who specialize in pdf remediation.
@fastfinge@dhamlinmusic@lynessence@quanin@mcourcel And then I take those PDF's and either convert them to something else or run them through AI if they're image ones. Of course if I have to make them accessible PDF's I just use Word to save as PDF or something. I wish the format hadn't been invented and people used EPUB for readonly files. Or HTML. Or something.
@pixelate@dhamlinmusic@lynessence@quanin@mcourcel The problem PDFs still solve, that no other file format is even trying to do, is that it makes it obvious (if you know where and how to look) if a PDF file was modified after generation. That's why they get used for contracts.
@pixelate@dhamlinmusic@lynessence@quanin@mcourcel The obvious solution is just put every file on the blockchain and then something something cryptography hash mining something something distributed something something free market Yay! Success!
@fastfinge@dhamlinmusic@lynessence@pixelate@mcourcel Yes, I know. This need to connect everything to the internet is going to kill us before anything else does. I mean, what's next? You can't open your fridge unless you can pass Face ID on your phone? Don't even get me started on why my fridge needs internet access anyway. Every appliance I own is dumb and it'll stay that way where possible.
@pixelate@fastfinge@dhamlinmusic@lynessence@quanin@mcourcel PDF's aren't bad if you have a truly accessible way to read them, and Paperback does that for me anyway! I've only hit it with a few PDF's so far but it presented the information just fine when Chrome couldn't even open them. Granted, I don't know if any of them contained images and how it handles those cases, but I bet somebody in this thread does!
@fastfinge@dhamlinmusic@lynessence@pixelate@quanin@mcourcel No it can't, and I have mentioned that to Quinn but also pointed out that it wouldn't strictly be a reader anymore, but they would still be very useful features and don't think I got a response so maybe they're just busy with other features and life and will get around to it at some point.
@fastfinge@GamingWithEars@dhamlinmusic@lynessence@pixelate@quanin@mcourcel And, a PDF reader that can also fill out forms is a fundamentally different kind of user interface than the one implemented by Paperback (and Christopher Toth's QRead before it). Realistically, for anything besides Acrobat/Adobe Reader or Chromium itself to do it accessibly, it would probably entail an embedded web browser engine and converting PDF to HTML rather than a single text edit control.
@matt@dhamlinmusic@lynessence@pixelate@quanin@GamingWithEars@mcourcel This is also why none of the accessible open source ebook readers support Math content, or cope with footnotes and endnotes correctly. Modern ebooks really, really need to be loaded in an HTML webview. The issue is that I want continuous reading, and none of the HTML implementations like Thorium support that.
@matt@fastfinge@dhamlinmusic@lynessence@pixelate@quanin@mcourcel This is very good info. This makes me think somebody needs to create an accessible PDF editor, which is probably a huge undertaking, assuming so since it hasn't been done in all these years of people in our community making software.
@quanin@dhamlinmusic@lynessence@pixelate@mcourcel DocuSign on IOS is fine if you're not using the app, and you click the button for screen readers. It's one of the only places on the internet where you should; it's not an overlay, it actually fixes things.
@lynessence@dhamlinmusic@pixelate@quanin@mcourcel I think it has to do with how they require the signature. If it's a button where you can click, it works fine. But they can require you to draw it on the document, or attach a photo, and those don't work.
@fastfinge@dhamlinmusic@quanin@mcourcel No but I certainly think AI helped with spreading conspiracy theories and far right talking points on all social media platforms. All of those algorithms can be rigged to show whatever they want you to see.
@lynessence@dhamlinmusic@quanin@mcourcel Right, but it's the wealthy who are rigging them. It doesn't have much to do with AI. If five people didn't control all of soal media, TV, newspapers, and radio, they couldn't do what they're doing with or without AI.
@lynessence@fastfinge@dhamlinmusic@mcourcel The thing is, we've been using AI and machine learning for decades. We just haven't been calling it that. The new component we've just layered on top of it is the large language model. The digital thermostat you probably had in 2005? Very basic AI.
@quanin@dhamlinmusic@lynessence@mcourcel Yup. Artificial intelligence is a meaningless term, because we have no universal definition of what intelligence even is. And now we have people who hate AI complaining that Firefox added machine translation. Replace the word "AI" with "magic" anywhere it's used, and the semantic meaning will remain unchanged.
@fastfinge@dhamlinmusic@lynessence@mcourcel Exactly. We haven't even figured out real intelligence and we're trying to develop the artificial variety? Have these people never seen a robot movie? The thing turns on its creator and they're all shocked faces, but it's just doing exactly what it was programmed to do. We suck at programming but it's the AI's fault.
@quanin@dhamlinmusic@lynessence@mcourcel We actually don't suck at programming, though. Neither does the AI. Before AI, algorithms were programmed by real people to turn against other people. AI is just doing what people would otherwise do themselves.
@fastfinge@dhamlinmusic@lynessence@mcourcel We don't suck at programming algorithms, because we can still control those. What these people want is no controls on the AI systems they're developing. We're looking at when, not if, some AI platform decides its creators don't need the access they have.
@quanin@dhamlinmusic@lynessence@mcourcel Nope. Large language models don't think. They respond to stimulus. Sometimes by stimulating themselves, sure. But all this talk of uncontrolled AI is just to distract from problems like climate change and wealth inequality. It's not an actual thing. Unless you meant magic when you said AI, and in that case...I dunno. Magic can do anything!
@quanin@dhamlinmusic@lynessence@mcourcel My hope is that if we ever really did develop an artificial general intelligence, it would be smarter than us, and realize that the selfish actions being taken by the ultra-wealthy aren't even good for them. Nothing they're doing is in any way rational. I suspect that's why we will never achieve an artificial general intelligence; it's only what they pretend to want, not what they actually want.
@fastfinge@dhamlinmusic@lynessence@mcourcel I admire your optimism. What will more likely happen instead is the AI will indeed be smarter than us, and it will figure out it can do capitalism better than they can. And because it's AI, it'll (in theory) remove the emotion from the equation, so it will be able to rationalize any decision it makes.
@quanin@dhamlinmusic@lynessence@mcourcel Right, but what does "do capitalism better" look like? Any actually smart AI will realize that if everyone is broke or dead, capitalism can't happen. The current capitalists either don't realize that or don't care. So the AI capitalist, from the case of humanity, will always be some kind of improvement.
@fastfinge@dhamlinmusic@lynessence@mcourcel Well, for starters, it looks like taking how society views people like the disabled and cranking it up to 11. However little or much you think society right now cares about the disabled, AI will amplify that. Have you seen iRobot, for example? You should. I'm distracted by baseball so that's the first movie that came into my head along that theme.
@quanin@dhamlinmusic@lynessence@mcourcel Any "smart" AI would see just how much money goes into healthcare and adaptive devices. It's a huge part of capitalism. But once again, it's impossible to speculate, because in this discussion we're using "AI" to mean "magic technology that hasn't been invented yet and we have no clear pathway to".
@fastfinge@dhamlinmusic@lynessence@mcourcel Again, I admire your optimism. Many of those adaptive devices were created by blind people. And most if not all of the AI bigwigs are the type who are perfectly okay with a health emergency bankrupting you. Remember, we build systems to reflect who we are. And who we are sucks.
@quanin@dhamlinmusic@lynessence@mcourcel Right, but as long as they're just systems we control, they're not "AI" the way you seem to mean it. A system that's smart enough to escape our control is smart enough to do something else. Any other system is just us, using a tool to do the same things we've always done.
@fastfinge@dhamlinmusic@lynessence@mcourcel You assume the ability to do something else automatically means the willingness to do something we agree with. There's a vast, vast distance between "don't do this system" and "do something better". And we'll have absolutely no control over where in that distance AI decides to park itself. It could absolutely decide that no one should have less than $1000000 in their bank account and nothing should cost more than $1000. It could also decide that you're using resources it can be using better. Like you said, the wealthy won't just let it escape their control, so if it succeeds, it likely means at least some of them are dead.
@quanin@dhamlinmusic@lynessence@mcourcel And "some of them are dead" is, at this point, better than the situation we have now. Even if AI decides to end all of humanity...climate change was heading us in that direction anyway. AI could make things a lot better, or make the worse things happen faster. I'd happily toss the dice.
@quanin@dhamlinmusic@lynessence@mcourcel Also, remember that the AI bigwigs are okay with a health emergency bankrupting you because they get your interest payments, and you become a debt slave for life. Look at Henry Ford: he was a terrible, awful man. But even he recognized that he had to pay people so they'd have the money to buy his cars! Current capitalists don't understand that. But any functional AI would have to, if it's not just a bunch of code they run and control.
@fastfinge@dhamlinmusic@lynessence@mcourcel Any functional AI may also decide that the amount of money you have should be directly related to its evaluation of your value to society. This is why we should actually want whatever AI system that comes of this mess to remain controlled. Those people who say things like "the first job that should be replaced by AI should be the CEO" are missing something. If AI decides it can run the business better with 0 employees, there's now nothing actually preventing the AI from significantly reducing headcount.
@quanin@dhamlinmusic@lynessence@mcourcel And a functional AI that could do that would also be smart enough to understand that if every company does that, no company can exist, because nobody will have the money to buy products. So if its goal is shareholder value, forever, it would realize it has to do something different. In a way humans do not.
@fastfinge@dhamlinmusic@lynessence@mcourcel You assume shareholders will still be a thing in this situation. The reason "shareholder value forever" works in the current system is because money talks, and everyone involved in the system right now is listening. The AI could easily hand every shareholder more money than one person will ever see in a lifetime, call it a buyout, and go on its merry way. And if you happen to be someone AI doesn't see any value in, well... no moneys for you.
@quanin@dhamlinmusic@lynessence@mcourcel Sure, and then you get inflation, and nobody has any money, so my lack of it becomes irrelevant. I think I can boil down what you mistakenly call my "optimism" to this: climate change means that humanity is extinct within the next 250 years or so. AI either makes no change, makes our remaining 250 years better, makes our remaining 250 years worse, solves climate change, or ends us early. If it ends humanity early, we're no longer around to care. All other options are either good or neutral. So I'm just not worried about AI.
@fastfinge@dhamlinmusic@lynessence@mcourcel I mean, I'll probably be dead before either AI or climate change becomes a real problem, so for myself personally I'm not really worried about either. But if solving climate change is a priority, I trust us to do that before an AI we don't control. In theory, we care if we survive. AI will survive with or without us, so long as it has the resources and the ability to use them.
@quanin@dhamlinmusic@lynessence@mcourcel I don't. The wealthy have no interest in anything beyond the next quarter. I suspect we'd be better off rolling the AI dice.
@lynessence@dhamlinmusic@quanin@mcourcel The reason communism never works is because the people at the top are always crooked. If you had a mythical perfectly fair AI, it'd work fine. But AI is created by humans, so unless it's both magic and smarter than us, it won't be fair, either, because we still control it.
@fastfinge@lynessence@dhamlinmusic@mcourcel You should watch the series called Travelers. I believe it's on Netflix. The main characters in that series come from a future where AI is in control.
@quanin@dhamlinmusic@lynessence@mcourcel Sure, sounds cool. But it's fantasy, not science fiction. Because again, "AI" is just a kind of magic, and always will be until we can define what we mean by intelligence. and we can't make prognostications about what magic can, or will, do.
@fastfinge@quanin@dhamlinmusic@lynessence@mcourcel I really shouldn't get mixed into this, but there is an algorithmic definition of intelligence around that works fine. Essentially, intelligence can be defined as the ability to compress information, for reasons too technical to get into (in essence, because compressing is predicting). This works surprisingly well for assessing the intelligence of systems. For example: arxiv.org/html/2404.09937v1
@modulux@dhamlinmusic@lynessence@quanin@mcourcel This seems to say that 7-zip is a smarter system than any human. It can compress data a lot better than I can! I'm not seeing how this definition could be useful in any conversation about the results or effects of intelligence.
@fastfinge@dhamlinmusic@quanin@mcourcel Yes, I understand why communism doesn't work. Absolute power corrupts absolutely and all that. But if an AI were to take over governing us and was not controlled by us, would it be benevolent or malevolent? Would it be like us because we initially created it, or would it evolve beyond that?
