log☇︎
713 entries in 0.904s
mircea_popescu: asciilifeform this discussion inspires me to the following ai standard : a computer that can distinguish code from data is rational.
mircea_popescu: farg is probably one of the better approaches to the ai problem, at any rate.
asciilifeform: at the time he actually was trying to scarf up all there was written re: ai, hoping to build something new
asciilifeform: mircea_popescu: the idea is that the mega-ai appears and has, somehow (forget physics for a moment) a total record of all things
mircea_popescu: "Roko's basilisk is a thought experiment that assumes that an otherwise benevolent future artificial intelligence (AI) would torture the simulated selves of the people who did not help bring about the AI's existence. [...] The concept was proposed in 2010 by contributor Roko in a discussion on LessWrong. Yudkowsky deleted the posts regarding it and banned further discussion of Roko's basilisk on LessWrong after it had ☟︎
decimation: but fear the great AI demon!
trinque: asciilifeform: yudkowski even invented the beast or satan or w/e, that AI that resolves to abuse your eternal soul forever if you oppose it
asciilifeform: much less anything actually hard like ai
asciilifeform: (did i mention his 'human-equivalent ai' was to be written in it? this is, of course, before he pronounced fatwa on anyone contemplating such a thing)
asciilifeform: sorta like the 'if it works, it's not AI' thing
thestringpuller: only LISP use in my undergrad program is in the AI courses
Vexual: *ai
Vexual: my ai prefers wrongness
adlai stumbles upon https://www.cs.cmu.edu/afs/cs/project/ai-repository/ai/lang/lisp/impl/alsp/0.html
asciilifeform: 1) ocr doesn't work because 99.99% is still ludicrously low - human reader -who fully understands the text- still required 2) no ocr exists or is reasonably likely to ever exist for anything rich in notation 3) the machine-readability of most mathematical work is a fully ai-complete problem
mircea_popescu: "oh, i also built natural language ai"
asciilifeform: the thread-will-meet-the-arse-correctly-but-without-foreknowledge is called 'intelligence' and folks call any solution that calls for it 'ai-complete'
the_scourge: asciilifeform: thanks. historically all i know about is really the AI lab stuff. i went through a phase where i read the jargon book and some more 'original' sources about the AI labs. but this was almost 20 years ago
mircea_popescu: exact fucking equivalent of the "expert systems" failed msword&excel-based approach to hard ai
mats: because 'hard ai' doesn't exist yet.
mircea_popescu: a well, so much for ai.
thestringpuller: i'd love to have asciilifeform AI controlling my house
thestringpuller: asciilifeform i hope they model an AI construct after you one day.
mircea_popescu: the reason this is even specced is to provide some sanity while everyone wastes their life in that ai quest.
decimation: well, the problem ai planes & ai soldiers is that the poorly wrought heuristics that went into the design would quickly be learned and exploited by the enemy
ascii_modem: ai soldier - aha
decimation: re: air war attrition << manned aircraft are on the way out anyway, with proper secure comms and a decent AI, a robot army of aircraft makes more sense
mircea_popescu: cazalla it's emerging ai huh ?
decimation: asciilifeform: and fearing the ai holocaust
undata: seems like the AI doom scenario is just the inverted singularity
decimation: so I find these AI doomsayers highly amusing. most of them would laugh at a Christian's claim of revelation (knowledge which comes neither through induction nor deduction) but are more than happy to jump on board the eliezer yudkowsky tard train because "I know it is true"
asciilifeform: mircea_popescu: mr y & friends believe in an ai equivalent of fermi's 'ignition of the atmosphere' hypothetical catastrophe.
mircea_popescu: perhaps this latter also different from a purely numeric ai
mircea_popescu: that's different from a mathematically bound ai, no matter how clever
mircea_popescu: they're thinking of a politically relevant ai.
mircea_popescu: who was that guy all worried about hostile ai ?
mircea_popescu: http://cdn.collarspace.com/photos/2036091.jpg << said ai.
mircea_popescu: asciilifeform incidentally, until such a time as natural language works for coding, there won't be the "evil ai" whatever guy's worried about.
mircea_popescu: hence well trained pilots, and training ai and so on
kakobrekla: since we are pasting links re ai; http://i1.kym-cdn.com/news_feeds/icons/original/000/015/610/48a.gif
decimation: I have an AI cult to sell
mircea_popescu: problem is i can build a machine with zero crosstalk and i won't have built an ai.
mircea_popescu: and yes, one of the many different reasons bayesianism is little more than "a cult of ai without moon or minsky aka stupid"
assbot: xkcd: AI-Box Experiment ... ( http://bit.ly/1xYdihM )
mircea_popescu: well it all started with you asking for a definition of ai.
mircea_popescu: otherwise windows is ai for crying out loud.
mircea_popescu: <mircea_popescu> any computer program of which identifiable components can be unambiguously named is not capable of displaying AI. << this obviously requires you to be allowed under the hood.
asciilifeform: this goes back to my complaint about the ai wishers. they aren't asking for 'intelligence', just a mechanical but recognizable version of themselves
mircea_popescu: so your argument is that ant meets my definition of ai ?
asciilifeform: (from a hypothetical 'ai')
mircea_popescu: i superficially suspect this discussion proves there can never be turing ai.,
mircea_popescu: well yeah but if it did violate restatement then either i'm wrong or no ai can be made to run on lisp machine.
asciilifeform: it's a humble tool, no ai to it at all
asciilifeform: what 'ai and stupidity' was about, was an actual pestilence of pseudoscience, where folks convinced themselves and others that they were accomplishing 'great things' while really pulling each other's dicks in a most stultifying way
mircea_popescu: any computer program of which identifiable components can be unambiguously named is not capable of displaying AI.
mircea_popescu: mp's lemma of artificial intelligence requires any computer program that exhibits in fact ai to depend in its functioning on the naming of its functions, and that self-metaprogramming be a part of its working.
asciilifeform: it's an epic paper, that not merely rained but pissed on the exuberant 'ai' parade of the '70s academe.
asciilifeform: http://home.engineering.iastate.edu/~alexs/classes/2011_Spring_585X/readings/02_Principles/AI_Meets_Natural_Stupidity.pdf ☟︎☟︎
assbot: Some AI Koans
assbot: Logged on 25-09-2014 21:26:13; mircea_popescu: really the proposition of making a living with trading bots is the response of the engineer who confronted with the fundamental impossibility of making money picking stocks decides to write a hard ai to do it.
mircea_popescu: bounce but one specifically optimised for the role of following you around taking pictures of things. like a sort of gopro with ai.
asciilifeform: mircea_popescu: guy learned russian phonetically, it sounds totally like ai << gets worse. 1) https://www.youtube.com/watch?v=PmAxM5GixMM 2) (danger! bagpipe) https://www.youtube.com/watch?v=Ep0LMkJBBbs
mircea_popescu: guy learned russian phonetically, it sounds totally like ai.
undata: it's like Elon Musk recently with AI. He's been calling that "potentially more dangerous than nukes"
jurov: something like that may pass after well done scare of "child raping AI" but I don't see how it can be enforced in practice
asciilifeform: re: elon musk: two thing happening. a fella named yudkowsky has been pushing for 'regulate ai, or it will destroy us all!' line for a little short of a decade now. with him, of course, to be the fuhrer. (see lenat discussion. he considers anyone trying to resurrect lenat's work, assuming it wasn't fraudulent - we don't actually know this - to be a dangerous criminal worthy of extermination.)
bounce: next we'll learn that the collective twits twittering on twittar causes AI
jurov: http://www.washingtonpost.com/blogs/innovations/wp/2014/10/24/elon-musk-with-artificial-intelligence-we-are-summoning-the-demon/ another AI expert
mircea_popescu: "i know, let's pay bob from Consultancy DerpAces LLC a million funbux so he can write an AI expert system that'll generate the text our readers would have put out! Save them the trouble!"
TheNewDeal: this an accurate translation, or is it butchered by AI - Wittgenstein , " Whereof one can not speak, thereof one must be silent
Vexual: ;;later tel xinxi whats the difference between an expert system and ai?
bounce: txtspk, bringing humanity into easy AI territory from the other end
Vexual: https://www.youtube.com/watch?v=YCjspyo-_aI
assbot: The Voice of the Turtle: Whatever Happened to AI? | Lenat | AI Magazine
decimation: ascii hopefully we can share the hard-ai-god's gasenwagen
mircea_popescu: i got the space, just, some derpy kid's idea of ai will prolly be to try and "loic" it
asciilifeform: now we can finally destroy the universe by enabling any kid to build rogue ai.
mircea_popescu: really the proposition of making a living with trading bots is the response of the engineer who confronted with the fundamental impossibility of making money picking stocks decides to write a hard ai to do it. ☟︎
mircea_popescu: one of the biggest promises of hard ai is the computer that slaps the human upside the head
moriarty: did he watch too much AI doomsday movies?
decimation: yeah I guess there's this fantasy that he can personally steer research toward non-angry-ai-god ai
mircea_popescu: "ETA: How concerned should we be that DARPA is going full steam ahead for strong AI? Perhaps not very much, given the failure of at least two of their projects along these lines"
jurov: Comment author: Eliezer_Yudkowsky This is a road that does not lead to Friendly AI, only to AGI. I doubt this has anything to do with Lenat's motives - but I'm glad the source code isn't published and I don't think you'd be doing a service to the human species by trying to reimplement it.
moriarty: with respect to AI
bounce: you're saying hard AI needs hardware with specific lisp support?
moriarty: i see copumpkin on here, he's also into AI :)
asciilifeform: moriarty: no. went to uni, ordinary 'cs', but vomited from 'ai' courses.
moriarty: asciilifeform, fair enough, still if you've formally undertaken AI as a major, that's something
mircea_popescu: so the idea is, the tool required (os/stack/machine) has no industrial utility and without convincing the industry it needs ai to sponsor the thing, there's never going to be either ?
moriarty: i recognise some AI practitioners on here
asciilifeform: mircea_popescu: the final doom of smbx corp. was sealed by 'ai winter.' in turn, the demise of the lm ensured that there would be no 'spring' - because carrying on the work in a non-homoiconic language is much like attempting algebra with roman digits.
moriarty: asciilifeform, and they didn't inform you that hard AI was a largely abandoned endeavour
decimation: asciilifeform: in the end it doesn't matter if the gasenwagen is driven by robot/ai or communist bureaucrats
moriarty: asciilifeform, everybody who ventures into AI informally makes that exact same mistake
asciilifeform: i used to be rather into the idea of 'hard ai' before realized that we don't even know how to build a computer.
moriarty: decimation, when i was studying and eventually doing research in AI, we got to use anything we like
moriarty: decimation, heh the modern AI researcher uses any language at his disposal
bounce: AI cure for gayness?
asciilifeform: decimation: you won't. because 'ai winter.' (and you also won't hear many people discuss how such 'winters' are the flip sides of 'summers', bacchanalia of grantsmanship and graft)
decimation: asciilifeform: one day I hope to find a modern 'ai researcher' who admits to have tried programming in lisp