There is a scene in "The Algebraist" (2004, Ian M. Banks) the leader of the invading space army (who is ruthless and petty) makes a demand for information of the gas giant aliens known as "the dwellers."
-
@futurebird The fact that we have built systems in which "sense" can be checked more reliably, and tested against specifications, is probably a reason why the main success use case for LLMs is semi-automated programming. But how much sense can we reliably make up in the air this way, and expect our systems to maintain coherence (internally, and to our expectations), without having made the work to really inspect it ourselves (and without having a deterministic process to do so, like a compiler)?
@futurebird But the main point is not so much about computers, it's about our brains, and how primed we are to see meaning where there is none, so when the message is really designed by a complex machine to really look like something with meaning, it's really, really, hard not to see any in it, if you pay a little attention to it. If you do, you have to go much deeper into it, to see the gaps, the inconsistencies, and we, as a species, are not as great as that as we think we are.
-
The exploration of AI we need is the one that grapples with the way that people will ascribe life, agency, trust to the obviously inanimate.
Think about the movie "Castaway" Chuck Noland (Tom Hanks) is so alone that he makes himself a friend/god out of a volleyball with a bloody hand-print on it. He talks to it. He prays. He needs it to limit his creeping madness in isolation.
@futurebird The sad fact is that fiction and character designers need to stop depicting machines as human and start depicting them as what they are: doppelgangers. Facestealers. Even (loaded with millennia of misogyny as they are) the original mythical depictions of succubi and sirens as man-eating monsters.
Because that's the truth of what these things are: human-form mimics that lure real people to their horrific doom without even the physical capability for morality or remorse.
-
The exploration of AI we need is the one that grapples with the way that people will ascribe life, agency, trust to the obviously inanimate.
Think about the movie "Castaway" Chuck Noland (Tom Hanks) is so alone that he makes himself a friend/god out of a volleyball with a bloody hand-print on it. He talks to it. He prays. He needs it to limit his creeping madness in isolation.
I humbly submit this essay from 2023:
https://ninelives.karawynnlong.com/language-is-a-poor-heuristic-for-intelligence/ -
The exploration of AI we need is the one that grapples with the way that people will ascribe life, agency, trust to the obviously inanimate.
Think about the movie "Castaway" Chuck Noland (Tom Hanks) is so alone that he makes himself a friend/god out of a volleyball with a bloody hand-print on it. He talks to it. He prays. He needs it to limit his creeping madness in isolation.
@futurebird
Blindsight by Peter Watts has some interesting exploration that feels relevant to this moment with LLM chatbots... -
There is a scene in "The Algebraist" (2004, Ian M. Banks) the leader of the invading space army (who is ruthless and petty) makes a demand for information of the gas giant aliens known as "the dwellers."
He proceeds to shoot living people, (just random ordinary people) out of his ship's gun like bullets to suffocate in space.
A decade ago I thought this was a little silly and over the top. "Come on Mr. Banks, I understand you want to lampoon warmongers, but this is too much."
I get it now.
@futurebird
I wonder if the Spielberg/Kubrick "A.I." deserves a re-watch? My memory of it is of how, watching it, I instantly wanted to be sucked in to it as a re-telling "Pinocchio", until I started realizing that the kid was just a toaster. The more unsettling and unwatchable it becomes, the better it got. -
@futurebird
Blindsight by Peter Watts has some interesting exploration that feels relevant to this moment with LLM chatbots...@futurebird
I'm thinking in particular of the alien(s) who take a snapshot of the Earth, then later we can have a conversation with it/them, but there's no "there" there, not conscious/sentient as we know itFireflies as LLM training, Rorschach as ChatGPT
-
@futurebird I love Banks’ Culture novels, and that society is closest to my sci-fi ideal, but I’m *very* dubious that humans could have much shared interests with miles-long AI-powered warships (however cool their names may be).
@michaelgemar @futurebird I’ve always thought the Minds treat humans as pets. I never had much of a shared interest with Twitter the Sugar Glider, but I would let her lick yogurt off my finger because she made such charming “this is *so* good” noises.
Jinx the Red-Eared Slider (turtle) became increasingly tiresome as he aged, but we couldn’t just throw him away. That’s not what a respectable person in my culture would do. Same for Minds?
-
@futurebird But the main point is not so much about computers, it's about our brains, and how primed we are to see meaning where there is none, so when the message is really designed by a complex machine to really look like something with meaning, it's really, really, hard not to see any in it, if you pay a little attention to it. If you do, you have to go much deeper into it, to see the gaps, the inconsistencies, and we, as a species, are not as great as that as we think we are.
@tshirtman @futurebird it's only a slight tweak on the Chinese Room at the end of the day. Humans communicate with each other in an embedded way, grabbing context and meaning from living in the same kinds of bodies. LLMs create a statistical reproduction of meaning from an unimaginable amount of data, comparable to Searle's filing cabinet. We just didn't believe such a filing cabinet could really exist, or that it would fool us.
-
@tshirtman @futurebird it's only a slight tweak on the Chinese Room at the end of the day. Humans communicate with each other in an embedded way, grabbing context and meaning from living in the same kinds of bodies. LLMs create a statistical reproduction of meaning from an unimaginable amount of data, comparable to Searle's filing cabinet. We just didn't believe such a filing cabinet could really exist, or that it would fool us.
IDK. To me the "Chinese Room" is about something else. Maybe the irrelevance of the inner-workings of a system. Maybe about how so much of our perception of "living" and "thinking" is tied to a particular pace of time.
This isn't the Chinese room, it's a magic 8 ball. But this magic 8 ball is the pastor of our church. Our savior and our guide and HOW DARE you disrespect him!
-
@tshirtman @futurebird it's only a slight tweak on the Chinese Room at the end of the day. Humans communicate with each other in an embedded way, grabbing context and meaning from living in the same kinds of bodies. LLMs create a statistical reproduction of meaning from an unimaginable amount of data, comparable to Searle's filing cabinet. We just didn't believe such a filing cabinet could really exist, or that it would fool us.
The LLMs are not "mad" ... the people who are using them in mad ways are.
-
@michaelgemar @futurebird I’ve always thought the Minds treat humans as pets. I never had much of a shared interest with Twitter the Sugar Glider, but I would let her lick yogurt off my finger because she made such charming “this is *so* good” noises.
Jinx the Red-Eared Slider (turtle) became increasingly tiresome as he aged, but we couldn’t just throw him away. That’s not what a respectable person in my culture would do. Same for Minds?
@marick @futurebird That’s a possibility, but it makes the Culture much less attractive.
-
IDK. To me the "Chinese Room" is about something else. Maybe the irrelevance of the inner-workings of a system. Maybe about how so much of our perception of "living" and "thinking" is tied to a particular pace of time.
This isn't the Chinese room, it's a magic 8 ball. But this magic 8 ball is the pastor of our church. Our savior and our guide and HOW DARE you disrespect him!
@futurebird @tshirtman that's Oz, right? You're talking about Oz. And so in Searle's story, there *is* no man behind the curtain. The wizard isn't a charlatan, instead he doesn't actually exist! We're just talking to a great head that echoes what other people have told it. We hear echoes that sound like answers. If people say that there is no wizard, we laugh it off or indeed get angry, refuse to look. Even if we agree there isn't any wizard, we may still say "the wizard told me"...
-
There is a scene in "The Algebraist" (2004, Ian M. Banks) the leader of the invading space army (who is ruthless and petty) makes a demand for information of the gas giant aliens known as "the dwellers."
He proceeds to shoot living people, (just random ordinary people) out of his ship's gun like bullets to suffocate in space.
A decade ago I thought this was a little silly and over the top. "Come on Mr. Banks, I understand you want to lampoon warmongers, but this is too much."
I get it now.
@futurebird Another Ian Banks fan? Yay, Team!
-
The exploration of AI we need is the one that grapples with the way that people will ascribe life, agency, trust to the obviously inanimate.
Think about the movie "Castaway" Chuck Noland (Tom Hanks) is so alone that he makes himself a friend/god out of a volleyball with a bloody hand-print on it. He talks to it. He prays. He needs it to limit his creeping madness in isolation.
@futurebird
Reminds me of a study I read years ago wherein researchers showed preverbal toddlers animated geometric figures “interacting” on a screen and reported that the youngsters reacted to the figures in ways that suggested they were ascribing agency and intentionality to the figures. -
There is a scene in "The Algebraist" (2004, Ian M. Banks) the leader of the invading space army (who is ruthless and petty) makes a demand for information of the gas giant aliens known as "the dwellers."
He proceeds to shoot living people, (just random ordinary people) out of his ship's gun like bullets to suffocate in space.
A decade ago I thought this was a little silly and over the top. "Come on Mr. Banks, I understand you want to lampoon warmongers, but this is too much."
I get it now.
@futurebird Banks got a LOT. I remember being horrified by some of the things he would write, and then looking around at the world and thinking that he might have been an optimist in some ways.
-
The LLMs are not "mad" ... the people who are using them in mad ways are.
@futurebird @meuwese @tshirtman
I equate LLM's and AI to the Rumanian Box. part of the scam was marks trying the box before the allotted time and assuming they'd broken it. Victor would explain how they'd "broken" it and sell them another one.
-
@futurebird I love Banks’ Culture novels, and that society is closest to my sci-fi ideal, but I’m *very* dubious that humans could have much shared interests with miles-long AI-powered warships (however cool their names may be).
@michaelgemarI always had a problem with his concept of AI's unilaterally plotting the course leaving it's captain to announce "Party!" to the crew. Can't remember which novel it was way back when. The one with a blonde Culture woman watching atop a dune ridge while her missile-knife decimates an oncoming army. The knife-missile idea to me was fascinating and science hadn't advanced enough to rule out it's conceivability. There's still the chance science could advance to a Type 2 Civilization. Will our cell phones become, a personel drone.
-
That inability to simply be alone is very real and very human. When you talk to a chatbot you are talking to a rubber duck, a volleyball, yourself.
But it isn't a self help exercise. It is a prescribed job requirement. It is a solution looking for a problem.
The "AI" SF story would not have amazing thinking computers who scare people who don't want to recognize they are human. It would have wooden dolls and people that get mad at you if you don't say "hello" and play along.
It’s late and I should go to bed, you just made me worry that all of my nice pocket friends are AI.
Although I did meet two of you IRL, so maybe I’m ok for now.
I should probably get more IRL friends.
-
There is a scene in "The Algebraist" (2004, Ian M. Banks) the leader of the invading space army (who is ruthless and petty) makes a demand for information of the gas giant aliens known as "the dwellers."
He proceeds to shoot living people, (just random ordinary people) out of his ship's gun like bullets to suffocate in space.
A decade ago I thought this was a little silly and over the top. "Come on Mr. Banks, I understand you want to lampoon warmongers, but this is too much."
I get it now.
Just finished that last week as an audiobook and posted the quote where Fassim first discovers that the twin Dweller is an AI and is terrified by being in a confined space with it. On the one hand, it's a lesson in breaking stereotypes. On the other, I'm not sure I believe the AI's claim that they were set up and were actually victims. It's a complex story and I may have missed it, but I don't recall a whole lot of reason to believe them.
It was definitely a departure from his Culture worlds where AIs are almost universally seen as benevolent. I find myself amused by the fact that I can suspend disbelief for faster than light travel and continent-sized orbiting space habitats, but have a much harder time believing in some future, super intelligent yet benevolent AI
️ -
Just finished that last week as an audiobook and posted the quote where Fassim first discovers that the twin Dweller is an AI and is terrified by being in a confined space with it. On the one hand, it's a lesson in breaking stereotypes. On the other, I'm not sure I believe the AI's claim that they were set up and were actually victims. It's a complex story and I may have missed it, but I don't recall a whole lot of reason to believe them.
It was definitely a departure from his Culture worlds where AIs are almost universally seen as benevolent. I find myself amused by the fact that I can suspend disbelief for faster than light travel and continent-sized orbiting space habitats, but have a much harder time believing in some future, super intelligent yet benevolent AI️I also noted, either in that post or in another one, that the way the Culture AIs talk sounds so much like the way current chatbots talk, that I can't help but wonder if our tech bro overlords were influenced by that when programming them.
Either way, to the extent that I ever have to deal with any of these robotic parrots, I'm going to take a cue from Culture characters and insult them by calling them "machine."
To be honest, I'm just waiting for the chance to say "shut up, machine," to a chatbot.
