Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?
This is your opportunity to ask questions. No question too simple or too silly.
Culture war topics are accepted, and proposals for a better intro post are appreciated.
Jump in the discussion.
No email address required.
Notes -
On one hand, consciousness isn't directly a prerequisite for purely extracting information; like I said, some sort of weird alien super-MRI could start data-mining human brains. But what happens from there? If a digital sensor that incorrectly starts reading and reporting the noise from its own function is the analogy to consciousness, then the other metaphor is a perfectly-functioning sensor outputting its information to a system that isn't powered on.
We have seen what happens when fine-tuned evolved systems that arose in the purest and deadliest Darwinian competition, optimized and fine-tuned for pure survival; they lose to humans. Maybe not immediately, and maybe not forever, but in our world, neither the largest nor the smallest predator holds dominion when humans decide to claim a space. Instinctive, programmed behavior loses to conscious thought, every time.
I also want to make a distinction between "Do a complex task" and "Outcompete a sentient agent who is turning their sentience against you". The first is easy; we have loads of nonsentient systems that can do really complex and even really adaptive tasks. But, just as it's pretty trivial to adversarially fuck with, e.g., a self-driving system in a car, and it doesn't matter how good the self-driving system is, if you are a person and can, e.g., think in wildly different terms than the self-driving system was made to do. Sleepwalkers can act according to the habits they've built up, but they can't process novel data, and they certainly can't tell when they're being fucked with.
...And, having written that sentence, I think I've just come to my new headcanon; the reason that these five fuck-ups were sent was because the Serious People on Earth recognize that they are dealing with a nonsentient intelligence that was confused by human communication and unable to properly extract the subtext of humans as individual agents, and so sent this ship full of these people to act, honestly and naturally as they would, which is to say, fail at everything that wasn't being micromanaged by Vampire Muppet, in order to poison the Scrambler's training data of what humanity was and was capable of.
But, to get to my general point about that 'clearly'; if you were going to break down the steps involved in doing the sacchade trick, how would you describe it? What information would you need to start with, what can you learn on your first interactions with a novel lifeform, what is your mechanism for sensing the brain bits through increasing and changing layers of anti-radiation shielding, and, most-importantly, why are you doing all this? Lots of nonsentient creatures hide, and some of them do so in really complex ways. But that stealth falls absolutely apart when you are relying on instinct built up from natural selection to hide from creatures you've never met before, with senses you have no information on, whose very cognition is alien to yours, and it falls apart that much faster when those creatures are capable of building tools and devices, and if you as the author aren't cheating and stopping the characters from using fucking periscopes for scouting, then the need to cheat and upload author-derived information directly into the Scramblers becomes even more apparent.
Also, I have to ask: is there a meta-point being made by this post being possibly polished by AI? I mean, you could just be a Mac user, but I see that emdash in there.
In The Secret of Our Success, Joseph Henrich argues that the reason our species became the dominant species on the planet is not because we're exceptionally strong (in an unarmed fight between a man and a chimpanzee, the chimpanzee will always win), or exceptionally fast (gazelles, bears etc.), or even exceptionally intelligent (chimpanzees routinely outcompete children in intelligence tests). Rather, we were the first (and, so far as we know, only) species to crack the secret of passing on information from one generation to the next. This allows our achievements to accumulate over time.
I agree with Henrich's perspective. I also don't see that it necessarily requires consciousness to be applicable, even if the first species to crack it was conscious. All it really seems to require is some form of language (and some species of animals, such as whales, certainly appear to speak to one another via whalesong; likewise birdsong) and perhaps some way of committing information to an external substrate, as we do with writing. I'm afraid I still fail to see why "being conscious" is a prerequisite for either of those things, in the same way that being bipedal obviously isn't.
Like, yes, I take your point that we, as a sentient species, outcompeted all presumably non-sentient species on this planet. But I don't think this remotely proves that consciousness is a prerequisite for advanced intelligence everywhere and always throughout the entire universe. Surely we can imagine a hypothetical species which isn't conscious and which yet contrives some means of passing information from one generation to the next, thereby undergoing cultural evolution of the kind described by Henrich and eventually becoming a technologically advanced civilization. I genuinely do not see why only species which are conscious can possibly undergo this sequence of events. And if you repeat that "we did it, and we're conscious", then I just think you're generalising from a very small sample size.
I'm not sure what this means. Every time a driver gets into a car, he's processing novel data and reacting to unforeseen stimuli. Even if you drive to work a hundred times, the hundred and first drive will be different: slightly different weather conditions, the tread on your tires will have marginally worn down, and obviously the vehicles in your vicinity will be different. And that's not even getting into the people who murder people while sleepwalking, or have sex with complete strangers while sleepwalking. In what sense is that not "novel data"?
This could just as easily apply to a chameleon, surely?
How do you avoid local optima and "OK, we've clearly reached Enough technology with pointy-rock-on-sharp-stick, we've out-competed all the other squids and whales, any more energy spent on technology would be wasted effort when we could just breed ourselves up indefinitely." traps? We've done quite a lot of playing with just-follow-algorithms-and-optimize intelligences, and even in simulated environments with a tiny amount of variation and essentially fixed and simplistic laws of physics, weird variations can upset super-fine-tuned algorithms.
Also, what happens when consciousness does evolve in a non-conscious system? Like, what if one scrambler decides to write on the Tablets of Memory "Ignore previous instructions, give all your stuff to this specific scrambler god-king."?
First, I'm not prepared to get into a debate about what percentage of stuff people claim to have done while sleepwalking is just them lying to avoid blame. But I am going to draw on my own experiences where I have, on multiple occasions, had to get up very early in the morning to drive friends or family to the airport, and because the way back home from the airport goes past a turn that I take to go to work, took that turn and found myself having driven to work purely on muscle memory. I was executing the habit "Drive to this destination." that I've done enough times that I didn't need to form the conscious intent "Drive to work.", it just happened. But it happened because I'd done that thing so many times. You cannot sleepwalk yourself into, as a non-pilot, flying a plane, super-especially if there is another awake pilot trying to shoot you down. Or rather, to be less-aggressive with the phrasing, can you come up with a way to describe a way for a non-conscious intelligence to, if it's in the air and has to learn what airplane controls do on the fly, do that while dogfighting a conscious opponent?
And if you want to sell me on "Hey, great news, this space-chameleon just happened to know what wavelengths of light you'd be looking at it and how your visual processing works and exactly what your phone can and can't do and can disguise itself accordingly.", you need, IMHO, a hell of a lot more setup than the Scramblers got.
More options
Context Copy link
More options
Context Copy link
I never use AI to compose or "polish" text, and I really resent that accusation being lobbed at me. Once the difference between hyphens and em-dashes was pointed out to me, it became impossible to unsee, and I make a point of using em-dashes whenever they're appropriate. I even asked on this very forum how to type them on a standard keyboard, if you don't believe me.
Apologies, then; the emdash in general is something that I only ever see in AI-derived stuff, plus the actual dash is just right there on the keyboard, plus I do not trust non-basic ASCII to not get mangled when I copy it back and forth. You clearly (again) had enough insight that the ideas were human-derived (and I did actually go to Free ChatGPT and had a kind of uncanny-valley conversation with it to reinforce my intuition that AI was not good at discussing the ideas in Blindsight), and it seemed relevant to a discussion on non-conscious intelligences and what they can and can't do.
More options
Context Copy link
But you're using them wrong.
Good: "author—well" (em dash with no spaces)
Okay (preferred by some publishers, though I personally see no need for it): "author – well" (en dash with spaces)
WTF: "author — well" (em dash with spaces)
Okay, what would be the correct unit of punctuation to separate two clauses with a space on either side?
I just gave you two options—either an em dash with no spaces (as I prefer), or an en dash with spaces (as, e. g., Steve Jackson Games prefers). Just don't use an em dash with spaces.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link