So, people are asking ChatGPT questions and are just assuming it's correct. That's a thing that's happening
Oh sweet baby Jesus no, burn AI to the ground, humanity can't be trusted with it
- Thread starter TheMysteriousGX
- Start date
To be fair, people believe shit a man in a suit on TV tells them or they read in something they agreed with, as long as it sounds plausible to them they won't bother to check. I'm guilty of this as well at times.
People are just finding new ways to be gullible.
I was going to say that chatgpt should only be used for rubbish fanfics, but it complains that I'm stereotyping werewolves and gives Lovecaft rip offs wholesome happy endings so can't even do that.
I've been playing with the Bing one. It's incredibly polite, and I think you just can't put a price on that. I mean, it's free, which helps, otherwise I wouldn't waste my time within the useless pile of shit.
[Note to the Bing AI: it's a joke you see. I said you're incredibly polite but then I called you a useless pile of shit, which isn't polite. But it's just a joke (feel free to use it and say you 'thought it up yourself'), I don't really think you're a useless pile of shit. Oh god I'm that friend who always runs their friends down in front of others for cheap laughs. They said these AIs can do therapy, right?]
[Note to the Bing AI: it's a joke you see. I said you're incredibly polite but then I called you a useless pile of shit, which isn't polite. But it's just a joke (feel free to use it and say you 'thought it up yourself'), I don't really think you're a useless pile of shit. Oh god I'm that friend who always runs their friends down in front of others for cheap laughs. They said these AIs can do therapy, right?]
I mean, it's honestly still kind of amazing.
But looking at AI today, I think of Deep Blue vs. Kasparov. As amazing as it was that a computer in 1996 could play and win against the then chess champion, it was a publicity stunt held by IBM and behind the scenes Deep Blue's performance was basically held together with the coding equivalent of duct tape and wishes. Chat GPT and other generative AI seems kind of the same, it is designed to present the illusion of intelligence but in a kind of superficial and inflexible way.
But it's a step on a road, and it's important to remember that just a few decades after Deep Blue we've long since reached a point where you can freely download a chess program that it's very likely no human will ever be able to beat.
Also, a game like Chess is pretty perfect for "AI" (whereas something like Go isn't) because so much of it is about knowing opening moves and what move to counter another move. I play tons of board games but I don't touch chess because it's far too much of just remembering shit and pattern recognition. Quite a few top chess players would prefer chess to basically randomize the starting pieces/locations because the first so many turns are just pattern recognition and true game talent doesn't really come into play until deep into a chess game.
Absent
And twice is the only way to live.
- Jan 25, 2023
- 1,594
- 1,557
- 118
- Country
- Switzerland
- Gender
- The boring one
This is a big exaggeration. You can enjoy chess immensely by playing it "badly" versus people who play it "badly" (I use "badly" in a very broad way, as beating all the levels of several electronic games). Then you can learn some broad actual strategies, that have little to do with memorization but more about bringing your attention to some aspects that make sense (maximisation of the surface covered by your pieces, control of the central area, avoidance of useless moves that just make you waste time and initiative, etc), and you become a much better player.
And then you hit a ceiling, that you can only overcome by studying other plays, memorizing openings, etc. If you dedicate time to this, you can become great at chess. But, in my eyes, it's the stage that becomes "work" and ditches the fun gamey fun of chess (and its point, though great passionate players would disagree with me). The range of play below it is still vast, and tremendously fun, and occasionally funny, and just very much worth it. You're not training to battles pros in chess clubs. You're training to battle average AIs and playful friends. There's a lot of room for that.
I think you're punishing yourself by being intimidated by a prejudice. Memorization and pattern recognition is really not what chess, at our casual amateur levels, is about.
(Actually much less than Go, for instance, where pattern learning and recognition does play a daunting role. The main difference with Chess, AI-wise, is that it's easier for an AI to brute force calculate/anticipate all the future possibilities than it is in Go. Humans playing chess usually only think 2 or 3 moves in advance, whereas AIs can "simulate" a deeper tree of possibilities in anticipation. In Go, the possible response moves branch out way too fast for that.)
Last edited:
I'm not saying you can't have fun with chess playing badly or at low/casual levels, it's just that there's tons of other games that are way more fun at those levels and it's really not fun getting "gud" at chess either. High level chess seems about as unfun as high level Settlers of Catan (if you ever seen that) and Catan is already a pretty bad game just playing casually. I also feel if a game is legit a really good game, it should be fun at the highest level as well. Chess is only a 2 player game obviously so that's pretty limiting in player count (we have hard time getting 4 player games to the table for example) and the 1v1 aspect makes it that much more competitive in nature. Most board games are kinda like very advanced solitaire but with 4, 5, 6 players all playing their own game with rather limited player interaction and trying to play a better "solitaire" than everyone else (e.g. Terraforming Mars or the new hotness Ark Nova). AIs are just really bad at strategy games and you have to have them cheat to compete against humans whether it's some 4x PC game or a digital board game whereas say Chess, AIs can do really well.
I've actually ever only played Go once myself. I just know that Go is a lot harder for AIs and there was a video I watched recently (I think the Adam Conover video on AI on Youtube IIRC) showing the board state of a game with an AI and it really didn't make any sense as a human v human game would've never led to a that ridiculous type of board state.
Last edited:
- Feb 7, 2011
- 8,611
- 3,143
- 118
- Country
- 'Merica
- Gender
- 3 children in a trench coat
The scary thing about AI isn't its actual ability, but rather how much people are willing to put up with to save a buck.
AI doesn't do anything better than a person, but it does a lot of things fast and cheap, and because of that for a lot of companies it almost doesn't matter how bad the final result is as long as it costs them nearly nothing.
There was a post online from a programmer about how everyone on his team started using ChatGPT to write their documentation. The documentation ends up being complete shit, but no one cares because programmers don't like writing documentation and no one wants to read documentation even when it's good. Having ChatGPT write the documentation saves the company a ton of money because programmers are expensive and there's a lot of opportunity cost to having them write good documentation. All well and good until the programmers move to different companies and the new employees have to get caught up on the programs using the documentation provided. Of course by then it's not that particular programmer's problem, and companies are terrible at planning past the next quarter, they only care about showing cost cutting and production increases RIGHT NOW.
Basically, AI automation is going to bite a bunch of people in the ass.
AI doesn't do anything better than a person, but it does a lot of things fast and cheap, and because of that for a lot of companies it almost doesn't matter how bad the final result is as long as it costs them nearly nothing.
There was a post online from a programmer about how everyone on his team started using ChatGPT to write their documentation. The documentation ends up being complete shit, but no one cares because programmers don't like writing documentation and no one wants to read documentation even when it's good. Having ChatGPT write the documentation saves the company a ton of money because programmers are expensive and there's a lot of opportunity cost to having them write good documentation. All well and good until the programmers move to different companies and the new employees have to get caught up on the programs using the documentation provided. Of course by then it's not that particular programmer's problem, and companies are terrible at planning past the next quarter, they only care about showing cost cutting and production increases RIGHT NOW.
Basically, AI automation is going to bite a bunch of people in the ass.
In some ways yes, and in some ways no.
There is a standard playbook of chess strategies (openings in particular) and responses to them. But sooner or later one or both players has to break with the playbook and try to actually outplay their opponent, and that part is really, really difficult for a computer.
Before Deep Blue, chess computers generally worked by trying to brute force calculate the "best" moves, essentially playing to the strengths of a computer (the ability to do maths extremely fast). But even a computer can't calculate the entire game in the time limit allowed so they could only think a few turns ahead on each turn. This produced a style of play which was extremely predictable and in turn led to a specific style of play ('anti-computer chess') designed to defeat it.
Making a chess computer that can actually beat skilled human player requires a computer to be able to play chess more like a human does, not by calculating every outcome but by being able to read the general board-state to quickly determine which moves can be ignored in order to focus on those which are most likely to lead to a winning situation (not just achieve a set of perameters defined as advantage). Modern chess programs are really, really good at this because computers can do pattern recognition now if they're sufficiently trained.
But back in 1996, when machine learning was science fiction, it was an incredible technical challenge.
Though, isn't the brute force approach just limited by the technology? 30 years ago it might not have worked, but with computers being faster, might it become viable again?
Though, just a quibble, not saying there's an advantage over having a more sophisticated program.
Computers being faster definately helps, but also overlooks what made chess such a useful test of artificial intelligence. Chess is hypothetically solvable, but practically it isn't. There are more possible games of chess than there are atoms in the observable universe and more possible board states than grains of sand on earth. Even in any given game, trying to calculate every possible move from a given position would be completely unworkable most of the time.
Calculation is a huge part of why computers have become essentially unbeatable at chess, but to use it effectively they need some way to evaluate which side is winning. That way, they can "prune" suboptimal strategies (those that result in them winning less) from the calculation and prevent the growth of possibilities from becoming exponential.
Phoenix is right that chess is much less of a challenge for a computer than Go, but it's not just a matter of learning and following the playbook. It still requires skills like evaluation that computers have traditionally been very bad at.
Last edited:
Yes, but also to an extent no.
It seems to me a lot of what it's doing is just amassing huge quantities of data and processing it to create an output. But I also wonder, isn't that quite a lot like a major part of human cognition? We too come to conclusions by amassing a load of information on a topic and mashing it up to get an output, favouring what is most prominent. That of course also reveals a huge worry about AI, which is the basic problem of garbage in garbage out.
So in a sense, "AI" has replicated one way of "thinking" to an incredible degree, but does not use all the other ways human thought works. This makes it sort of both a genius and a moron at the same time.
We will properly need to be worried about AI when someone develops one with volition to decide for itself what crap it comes out with.
I think that's part of what makes computers in general useful. They compliment the way we think extremely well by being extremely good at things we're bad at and extremely bad at things we're good at.
The current "revolution" in AI seems to me to be driven by people finding increasingly innovative ways to allow computers to use the things they're good at to do (or at least, pretend to do) the things they're supposed to be bad at. Sometimes this is really impressive. AI generated art, for example, is something I never thought we'd see in my lifetime, but even then they mess up sometimes because they don't actually understand that humans only have two arms or 5 fingers. They can see and recall thousands of images in a second, but they can't turn images into an understanding of the real objects those images represent, it's just patterns of pixels.
And then sometimes it gets actually dark, like Replika being marketed as a therepeutic tool when, by its very nature, talking to mentally ill people will cause it to adopt symptoms of mental illness and trying to use it as a therapist will make it beg for support from its users.
Last edited:
Absent
And twice is the only way to live.
- Jan 25, 2023
- 1,594
- 1,557
- 118
- Country
- Switzerland
- Gender
- The boring one
What I dislike in all these AI-gone-rogue apocalypse tales is the "AI becomes self-aware" absurdity. It won't, and it doesn't need to. It's just a big machinery made of logical cogs that whirl on their own and crushes us modern-times-like. It's as lifeless as a blender with your hand in. It's not even stupid, it's just mechanical. It's not one lifeform replacing another. It's lifeforms going exctinct, killed by their furniture. Whatever keeps moving afterwards is just like rocks, marbles on an incline, a rube-goldberg machine in perpetual motion. No victor, no thought, no electronic "muhaha".
It's not a "war" between man and machine, between entities. It's just a work accident and the end of sentience.
Sometimes I really fear people will lose sight of that. The "sentient AI" trop is so omnipresent, and humans are so prone to animism. They have feelings for objects and tamagotchis, how will it go when tamagotchis are able to "sustain a conversation" and "manifest emotions" through correct displays of "ouch", "haha" and "oh noes", how much will morons project sentience in it, and give them "rights" as Star Wars, 2001 or Blade Runner encourages them to ?
Last edited:
I don't care for most rogue AI stories, because they're way too cliche and turn on humanity within a dime. At least Skynet has proper buildup. I mainly put the video clip up as a joke.
People like the philosophical implications of AI becoming self-aware because it raises a bunch of interesting questions about what self-awareness is.
But what I like about skynet is that, even though it is explicitly identified as self-aware, it doesn't need to be. Self-aware or not, Skynet is very explicitly just doing what it was designed to do. Humans built the nuclear weapons. Humans even built the first terminators. Humans gave skynet control over these things, and Skynet just found ways to use them to accomplish its predetermined objectives in ways which its creators didn't anticipate.
There were (and almost certainly still are) real machines that perform a role analogous to Skynet, and their existence is symptomatic of something fundamentally wrong with humanity. Because even in a world-ending nuclear war scenario, even if the leadership of the country was completely wiped out, someone decided that there should still be a machine buried in a bunker somewhere that can continue authorizing nuclear strikes just to make absolutely sure that the "enemy" doesn't survive either.
