4 ms·
Without the clarification I would have assumed you were objecting to the idea of an AI "inventing" something, which I would agree with. I would have had no idea
by djur 4y ago
Without the clarification I would have assumed you were objecting to the idea of an AI "inventing" something, which I would agree with. I would have had no idea you were suggesting that children consuming AI-generated stories is obviously harmful, which I disagree with. So I appreciate the clarification.
Children are delicate in some senses but incredibly resilient in many others. Hearing a robot read a generic barnyard animal story wouldn't crack the top 50 most harmful experiences most kids have in a week.
- Jensson 4y agoI think they are afraid of indoctrination from Amazon based stories, not that children can't handle strange animal stories.
- mdp2021 4y agoAlso. About the «strange animal stories», more in detail (apart from my other comments): a story transmits a number of mental tracks and developmental inputs - from quality inherent in activities to structures inherent in the world, with parallel meta-justifications for its form ("telling this to suggest that"). Speaking of strange, let us remember the dall-e images that mix credible features and completely hallucinated unstructured blobs (e.g. "top shelf: soap, mirror, towel, alien excrescence"). Questions: why is that specifically, why is that complexively; what justifies that specifically, what that complexively... Intelligence. That model, translated into stories, would be "teaching oniric randomness" - plausible at low attention level but recognizable through detail and global scope.
- mdp2021 4y agoAnd just because children are resilient you would feed them sh*te? You are what you eat. If something deserves quality, it's children. Already "normal" stories are not trivial. Already the Märchen compiled by Jacob and Wilhelm Grimm do leave their specific track, which is not at all solar. Fortunately, the child may have many more books and sources to balance. But in the case of the hallucinated inventions of current ANNs, what would be the purpose - to distract the child, to waste its time, at that stage?! Surely, "not critically harmful" should not equate to "advisable".
- djur 4y agoI am indifferent to this specific phenomenon because it is a drop in the bucket. Frankly, a child would be better off hearing a story from a robot than watching many examples of children's TV or online video programming. I'll start caring about Alexa storytime once kids aren't spending almost half their waking hours looking at a screen, maybe.
- mrguyorama 4y agoEverything being shit is not a defense for a specific thing being shitty.
- djur 4y agoI mean, at this point I'm not sure what I'm arguing against, because I'm not quite following the English in this person's comments. Which is fine, my German would be much worse. But the impression I get is that they're treating these AI stories as something like a cognitive toxin. From my perspective the worst harm that's likely is that they'll be boring. Even calling it "shitty" is giving it more significance than it merits.
- mdp2021 4y agoSorry for the delay. > cognitive toxin I already regard much human intellectual production as «cognitive toxin» - and humans have the "modules" (I'll try to explain), even when underdeveloped. Now imagine that instead of a "human with them modules, whatever their state of development", you have a I/O box that inside has no actual intelligence (a critical "module"¹), no critical assessment of the quality of the output in all the required dimensions. (¹ "Intelligence" is actually more like a meta-module in this case, keeping more modules in check.) It may compose a story that, mixing elements from its "learning material" that it blindly judges coherent, mixes typical and random. The real elements of the story, not those of the facts but those of the purposes, will be missing from the parallel structures. When a genius composes, all the structures are checked. When an unintelligent thing composes, those coded-in will be implemented and the others will be easily deliria, with disconnected crumbles instead of structures. A story is not just a succession of events, but a structure of meanings. ...The child, being Intelligent, will notice that and go "WTF?!". Or absorb monsters instead of intellectual products. A text strong on the structure of "extended grammar of common language", but weak on the structures behind it, is «cognitive toxin», yes. The composer has to be Intelligent and see if there are faults - not just /formal/ faults. If an AGI had two parts judge and improve photorealism in a picture, it may work on making the building more credible, and be oblivious of the purposeless randomness of the situation of Julius Caesar helping Nixon roasting a skyscraper for a picnic of beavers depicted in (that this rarely appears in an automated composition because of the rarity of the links means nothing - a chain even of frequent events remains a random chain. And even a typical chain of frequent events would have no purpose per se). A text which appears legitimate at a shallow regard may easily be a monstrosity when you consider it attentively. Make an experiment: have a composer complex-ANN produce a story; read it and ask yourself (e.g.): "What does it teach you"? (This question does not exhaust the mess that could be behind a delirious composition, but is already an important indicator.) > the English Admittedly, I treat the language in a possibly uncommon way: I force it around the ideas I want to convey. I compose top down, not bottom up (not starting from the language but starting from the ideas and structures).