[HN Gopher] Why Mastering Language Is So Difficult for AI
       ___________________________________________________________________
        
       Why Mastering Language Is So Difficult for AI
        
       Author : TeacherTortoise
       Score  : 54 points
       Date   : 2022-10-07 16:43 UTC (6 hours ago)
        
 (HTM) web link (undark.org)
 (TXT) w3m dump (undark.org)
        
       | noobermin wrote:
       | Thankfully someone is pushing against the hype, good for him.
       | Thank god.
       | 
       | It really frustrates me when people, particularly people on here
       | specifically respond to criticism of NNs and the like by saying
       | nonsense like "but YOUR brain is a neural net!" as if all your
       | brain is but mimickry after all and there's nothing else. There
       | is this need to reduce all the mind's complexity down to a model
       | that you seem to understand, just because you understand it,
       | which seems like such a lazy, reductionist way to think.
       | Reductionism in general has value but it tends to narrow the
       | mind.
       | 
       | I feel like Gary Marcus is right, you don't just need the
       | mimickry even if it plays a role, you do need "symbolic"
       | understanding somewhere, and that makes it easier to determine
       | whether Trump is president vs. Biden at the moment, as they talk
       | about in the article. You do need concepts (symbols) deep down
       | somewhere, it seems so bizarre to think you don't because no one
       | even in their own lives lives that way or thinks that way. I mean
       | literally right now, me reading and writing comments on hacker
       | news is all about ideas and concepts, symbols in the
       | philosophical sense. I feel like even if it started as mimickry
       | when I was an infant, eventually the mimickry takes a life of its
       | own, I figure out _concepts_ themselves, and I no longer even
       | work in mere mimickry, I work with symbols.
       | 
       | It reminds me of Baudrillard's concept (unintended pun, I think)
       | of simulacra, it's not that simulacra are just representations of
       | the real and are thus fake, like money represents some amount of
       | food or gold or something and so money is fake. Money, as we all
       | know, is in fact very real! Our daily lives are shaped around it,
       | they shape our reality, things like loans, credit, bank accounts,
       | stock markets eventually become things we care about. Money is
       | therefore hardly fake. Eventually the simulacra develop a reality
       | of their own, in fact, they become hyperreal.
       | 
       | Ideas are the same, and any real AI will need them.
        
         | everly wrote:
         | Oh, he _loves_ to push against the hype alright. To his credit,
         | he's very smart - I think he sold an AI startup to Uber several
         | years ago. And I agree with a lot of his criticisms and general
         | AI philosophy.
         | 
         | Still, it's impossible for me to shake the feeling that he just
         | likes to hear himself talk and doesn't want to see progress in
         | AI.
         | 
         | I think I trust him more than Sam Altman though, not that
         | anyone's being asked to make that judgement.
        
           | ripe wrote:
           | > Still, it's impossible for me to shake the feeling that
           | [Gary Marcus] just likes to hear himself talk and doesn't
           | want to see progress in AI.
           | 
           | He says he wants to see progress in AI. He wants people to
           | get their heads out of the neural network cul-de-sac and use
           | other tools, too. His quote:
           | 
           | "Imagine a world in which iron makers shouted "iron," and
           | carbon lovers shouted "carbon," and nobody ever thought to
           | combine the two [to make steel]; that's much of what the
           | history of modern artificial intelligence is like." [1]
           | 
           | An adjustment will happen when the limitations of these
           | purely statistical methods play out and burn a few self-
           | driving car investors. Unfortunately, it will also probably
           | taint the entire field of AI and start another AI winter for
           | a decade or so.
           | 
           | [1] https://nautil.us/deep-learning-is-hitting-a-wall-238440/
        
             | everly wrote:
             | Thanks, that's a worthy addition (and good example of what
             | I mean about agreeing with his philosophy). I was probably
             | a bit too uncharitable in saying he doesn't want to see
             | progress.
        
       | lvxferre wrote:
       | [taunt]Most things that I hear about natural language processing
       | boil down to a bunch of codemonkeying assumers trying to
       | oversimplify linguistic phenomena that they are completely
       | ignorant about, and yet trying to sell it as if it was the next
       | best thing after sliced bread.[/taunt]
       | 
       | Anyway. There are three things that NLP is notoriously bad at:
       | 
       | 1. Using world knowledge to interpret utterances in a given
       | context. The article itself provides a neat example of that with
       | United-Statian presidents.
       | 
       | 2. Contextualisation that goes beyond the text that the utterance
       | is found in.
       | 
       | 3. Assessing the pragmatic purpose of an utterance.
       | 
       | But of course, "linguistics is useless for NLP!", the codemonkeys
       | say.
        
         | Der_Einzige wrote:
         | No, it unironically IS useless. You're welcome to bark up the
         | symbolic tree along with Chomsky, Gary Marcus, and others who
         | are about to fade into obscurity.
         | 
         | Or you can embrace the bitter lesson, which is that symbolic
         | techniques are considered harmful, like the future winners of
         | the AI race have:
         | http://www.incompleteideas.net/IncIdeas/BitterLesson.html
        
       | sbierwagen wrote:
       | Saved you a click: it's an interview with Gary Marcus.
        
         | CharlesW wrote:
         | Saved you another click if you also have never heard of this
         | person: https://en.wikipedia.org/wiki/Gary_Marcus
        
       | wpietri wrote:
       | It's nice to see some coverage of AI that is realistic about the
       | issues.
        
       | freeopinion wrote:
       | Mastering language is difficult for humans. I don't think most
       | people realize how often they just guess at meanings, and how
       | often they are wrong. Especially in spoken language.
       | 
       | Constantly asking for clarification is exhausting and gets you
       | labeled as pedantic and "difficult". So most people don't do it.
       | I'd guess that a fairly high percentage of human communication
       | results in a misunderstanding. This is largely masked by human
       | inaction or ability to disguise their lack of understanding. The
       | AIs have no ego and no incentive to hide their confusion.
        
       | kelseyfrog wrote:
       | Replace GPT-3 with Bob, language model with person, and deep
       | learning with learning, and you arrive at the conclusion that
       | we're not able to determine whether people actually understand
       | the world around them. Any utterance is indistinguishable from a
       | sufficiently advanced computation model which simply produces the
       | next token of text. Which is to say that the essential
       | discriminating characteristic isn't the fact that GPT-3 runs on a
       | computer, it's that it doesn't have a basis of reality rooted in
       | the present.
       | 
       | The article correctly recognizes this, but then fails to actually
       | make any meaningful contributions. Say the model was in fact
       | conditioned on the basis of a present reality, would that change
       | the interviewee's mind? The fact that it's difficult to say
       | simply hints that the interviewer didn't really do that great of
       | a job drilling down to the interesting questions.
        
         | beezlebroxxxxxx wrote:
         | I think implying that GPT-3 and a human learn a language in the
         | same way is an error. AI "learning" is very rigid; while human
         | learning is often a complicated mishmash of behaviour, social
         | feedback, reasoning, and imagination.
         | 
         | > conditioned on the basis of a present reality
         | 
         | If we grant this hypothetical, we might find what it "says"
         | fundamentally alien to us --- we might not even know it is
         | saying anything at all.
        
           | Tijdreiziger wrote:
           | Neural networks are modeled after the neurons in human
           | brains, so by definition, the way NNs learn approximates the
           | way humans learn.
        
             | kelseyfrog wrote:
             | How does the brain do backprop?
        
             | Der_Einzige wrote:
             | So, that's pretty wrong.
             | 
             | Spiking neural networks are possibly closer to being
             | biologically analogies, but they kind of suck to train and
             | don't have good performance.
             | 
             | It takes a lot of computer "neurons" to simulate a single
             | human neuron, and we can't simulate it perfectly in any
             | case.
        
             | yyyk wrote:
             | Neural networks were modeled after _a model_ of the neurons
             | in human brains, a model which we know by now is inaccurate
             | and misses a lot of details.
        
         | Jensson wrote:
         | > Replace GPT-3 with Bob, language model with person, and deep
         | learning with learning, and you arrive at the conclusion that
         | we're not able to determine whether people actually understand
         | the world around them.
         | 
         | Sure we can, there are diseases which greatly hampers your
         | ability to understand the world but doesn't hamper your verbal
         | abilities at all, Williams syndrome for example. They are great
         | at generating nonsense stories that doesn't make sense. GPT-3
         | is like that but much worse.
         | 
         | https://en.wikipedia.org/wiki/Williams_syndrome
        
           | chongli wrote:
           | For a particularly striking example of the opposite problem,
           | see people with Wernicke's Aphasia [1]. Here's an example [2]
           | of a man who had a stroke that left him with the disorder. He
           | can speak fluently but his speech is meaningless.
           | 
           | Byron can understand what's happening in the world around him
           | and can communicate it through body language (just see the
           | warmth in his eyes and his smile) but when he speaks
           | everything comes out in a stream of nonsense.
           | 
           | If we tried to apply a language-model-centric view of
           | intelligence we would rate Byron is unintelligent. That would
           | clearly be a mistake, as anyone can plainly see that he is
           | intelligent.
           | 
           | [1] https://en.wikipedia.org/wiki/Receptive_aphasia
           | 
           | [2] https://www.youtube.com/watch?v=3oef68YabD0
        
           | kelseyfrog wrote:
           | Right, so there is a characteristic other than computational
           | substrate that allows us to differentiate between systems
           | that have this ability and those that don't.
           | 
           | Say we apply the diagnostic criteria of William's syndrome to
           | language models and deduce that they meet the criteria for
           | diagnosis, then the diagnostic itself ie: the test results
           | are the grounding for the conclusion not the computational
           | substrate. In that way the discussion surrounding the
           | computation substrate is a red herring or at the very least a
           | unclear and roundabout way of grouping of the characteristic
           | we're actually interesting in.
        
         | mistermann wrote:
         | > Replace GPT-3 with Bob, language model with person, and deep
         | learning with learning, and you arrive at the conclusion that
         | _we 're not able to determine whether people actually
         | understand the world around them_.
         | 
         | Wouldn't the fact that people's facts do not match up prove
         | that they do not?
        
         | hinkley wrote:
         | I recall reading some advice to parents not to freak out if
         | they discover their child lying at an early age. The sentiment
         | was that lying is built on a mental model of another person,
         | and therefore requires some substantial development of
         | emotional intelligence to even make the attempt.
         | 
         | There's some romanticism in common between magicians, con
         | artists, and fraudsters about how unclearly 'normal people'
         | perceive the world so perhaps there's some truth to what you
         | say. And the perpetual problem with artists and visionaries is
         | that they see the world before everyone else does. For the one
         | group they see things as they are now while everyone else seems
         | to see things as they were 2, 5 years ago. For the other they
         | see what the world could be, by exercising the tools near at
         | hand.
         | 
         | One wonders then how an AI that actually understand things
         | would be seen. As a trickster? A misunderstood intellectual? A
         | madman?
        
       | benlivengood wrote:
       | Language is a communication method evolved by intelligent beings,
       | not a (primary) constituent of intelligence. From neurology it's
       | pretty clear that the basic architecture of human minds is
       | functional interconnected neural networks and not symbolic
       | processing. My belief is that world-modeling and prediction is
       | the vast majority of what intelligence is, which is quite close
       | to what the LLMs are doing. World models can be in many
       | representations (symbolic, logic gates, neural networks) but what
       | matters is how accurate they are with respect to reality, and how
       | well the model state is mapped from sensory input and back into
       | real-world outputs. Symbolic human language relies on each
       | person's internal world model and is learned by interacting with
       | other humans who share a common language and similar enough world
       | models, not the other way around (learning the world model as an
       | aspect of the language itself). Children learn which language
       | inputs and outputs are beneficial and enjoyable to them using
       | their native intelligence and can strengthen their world model
       | with questions and answers that inform their model without having
       | to directly experience what they are asking about.
       | 
       | People who don't believe the LLMs have a world model are wrong
       | because they are mistaking a physically weak world model for no
       | world model. GPT-3 doesn't understand physics well enough to
       | embed models of the referents of language into a unified model
       | that has accurate gravity and motion dynamics, so it maintains a
       | much more dreamlike model where objects exist in scenes and have
       | relationships to each other but those relationships are governed
       | by literary relationships instead of physical ones and so
       | contradictions and superpositions and causality violations are
       | allowed in the model. As multimodal transformers like Gato get
       | trained on more combined language and sensory input their world
       | models will become much more physically and causally accurate
       | which will be reflected in their accuracy on NLP tasks.
        
         | ripe wrote:
         | You might be right that these "multimodal" transformers, by
         | integrating additional data from non-text sources, would be
         | more capable than GPT-3. But I don't think that invalidates
         | Gary Marcus's point.
         | 
         | The word "model" is another of those words that Minsky called
         | "suitcase words"--- they can be used to mean many things. I
         | don't think Marcus is saying that that LLMs have "no model",
         | just that they don't have a model of the type that a symbolic
         | system could have. He gives many examples of deductions that a
         | symbolic AI system can easily do, which GPT-3 is simply
         | incapable of.
        
           | benlivengood wrote:
           | To hint at what I fundamentally mean by model I'd be
           | interested to see a symbolic model of vision. E.g. take 512
           | _512_ 3 numbers and give them names and then follow some
           | rules to arrive at "cat" or "dog". Image recognition, I
           | think, is demonstrably not symbolic. Likewise most
           | transformations from the real world to model state are not
           | symbolic. Within model state, symbolism may have uses but
           | Church-Turing claims that it isn't _necessary_.
           | 
           | It seems clear to me that if the CLIP-like part of Imagen or
           | Stable Diffusion can take an image made of pixels and yield
           | "cat" and similarly take "dog" and produce a 3D neural
           | radiance field that we can light just like any other 3D model
           | and recognize as a dog then there must be an accurate and
           | useful model of both how vision works and what dogs look like
           | and the English relationship between those two things inside
           | the machine.
           | 
           | I also wish Gary Marcus was replying to Google's Minerva
           | paper instead of GPT-3. The ability to answer multi-step
           | symbolic problems is basically here.
           | https://ai.googleblog.com/2022/06/minerva-solving-
           | quantitati...
           | 
           | I'd also note that decades of attempts at automated theorem
           | proving with symbolic systems haven't yielded similarly
           | impressive results. Now we have deep learning models helping
           | with proof search.
        
         | hinkley wrote:
         | Are you saying that GPT-3 doesn't grasp Object Permanence?
        
         | fspeech wrote:
         | Well put on what language is. Language encodes the delta of
         | worlds. There are way more unsaid than what is said explicitly.
        
         | noobermin wrote:
         | Reading things like this on hacker news and twitter reminds me
         | a lot like particle physicists unwilling to grapple with lack
         | of evidence for SUSY. Every time someone points it out, these
         | guys merely recite the holy texts (describe the model in gross
         | detail with more magic words as if the criticism is because the
         | critic don't understand it well enough), as if that addresses
         | the critique at all. They then eventually gesture at the energy
         | horizon, suggesting that "with more _s_ [0] you'll eventually
         | see my particles."
         | 
         | It's feels a bit similar here. There are clear limitations to
         | AI by just NNs, namely they like Stable Diffusion must be
         | trained on so much data, far more than any single artist will
         | ever see, and they still fall short sometimes. With NLP it
         | seems a little more obvious because the limitations seem to lag
         | actual speech understood by humans still, even with the massive
         | dataset again beyond what any single human will ever digest.
         | There are clear issues here, merely reciting to critics what
         | your model of intelligence is doesn't address the shortcomings
         | these systems have which any literate person can perceive. And
         | gesturing at the data horizon (you need even _more_ data??) is
         | not convincing either.
         | 
         | Apparently this cat (the interviewed author) has some baggage,
         | but the things he's talking about (symbols) have already been
         | studied and used by philosophers and linguists long before AI
         | or computers were a thing. It helps to build upon what other
         | researchers have already done, I don't know why NN researchers
         | don't seem to map what they are doing to existing research on
         | intelligence in other areas, namely cognitive science. Do NN
         | researchers have any contact with linguists or cognitive
         | scientists generally?
         | 
         | Again, being a little ignorant of his baggage, I do think a
         | hybrid approach sounds like it would be promising. Symbolic AI
         | failed in the 70s, NN is getting a bit of the way there now and
         | is pretty useful but misses a few things, why go all in on just
         | one type of approach?
         | 
         | [0] _s_ is basically the center of mass energy, what particle
         | physicists mean when they say the  "energy" of an interaction
         | (collision)
        
           | xwolfi wrote:
           | You say far more than any artist but I m not so sure. When we
           | spend 20 years looking at stuff before doing one passably
           | interesting work of art (bar absolute outlying geniuses),
           | it's not a small time nor a small training cost. It may feel
           | easy for us but our brain never stop ingesting information
           | for decades before being original enough.
        
           | Der_Einzige wrote:
           | "Every time I fire a linguist, the performance of my speech
           | recognizer goes up"
        
           | hackinthebochs wrote:
           | Yes, and the history of philosophy is replete with examples
           | where the learned opinion on a subject stymied progress.
           | Respect for one's intellectual forebearers is good, but being
           | constrained by them has proven to be detrimental in areas
           | where the learned opinion is largely speculative. The
           | progress made by deep learning is so unlikely anything that
           | came before, trying to shoehorn old methods and ontologies
           | into the new paradigm would only harm progress.
        
           | 317070 wrote:
           | This NN breakthrough in language is only about 2 years old
           | now [0].
           | 
           | People are working on your symbolic logic hybrid approach,
           | but not much interesting has happened yet. This is a bit
           | expected, since the short amount of time, and the fast
           | scaling these language models have done in those last two
           | years. It's been hard to outscale them with any other
           | approach, let alone by the particular one of symbolic logic.
           | 
           | I think the main discripancy with particle physics, is that
           | every iteration of bigger language models are also
           | significantly better, no diminishing returns in sight.
           | 
           | Better for concrete applications as well. There is therefore
           | no need to claim these models are scaled for scientific
           | reasons in the search of AI, the accountants are happy as it
           | is.
           | 
           | [0] https://arxiv.org/abs/2005.14165
        
       | gojomo wrote:
       | While language seemed difficult for AI for a while, it now seems
       | to be falling quite rqpidly to recent techniques.
       | 
       | (It reminds me a little of the early years of aviation. In 1901
       | the US Navy's chief engineer called manned flight a "vain
       | fantasy", & in 1903 the New York Times suggested flying machines
       | would be possible in "one million to 10 million years". But it
       | was accomplished later in 1903, with rapid improvement after
       | that. Things that seem impossible now can become humdrum in a few
       | decades.)
       | 
       | What are the current best results from Marcus, or others using
       | his preferred approaches?
        
       | pmontra wrote:
       | TL;DR "A parrot's not a bad metaphor, because we don't think
       | parrots actually understand what they're talking about. And GPT-3
       | certainly does not understand what it's talking about."
       | 
       | BTW a parrot knows a big deal about the world. Not much about
       | human language but still more about language than GPT-3. Of
       | course a 737 doesn't know how to fly but it still flies people
       | around the world. Landing on a branch, not much.
        
         | Gibbon1 wrote:
         | Grandmother mentioned her aunt had a parrot. One time she and
         | her aunt are hanging out. There was a lineman up on the pole in
         | front of the house and the bird was curiously watching him.
         | After a bit a young lady walks by and the bird chuckles and
         | gives out a loud wolf whistle. And the lady then begins to cuss
         | out the lineman up on the pole. The bird giggled and then
         | giggled off and on for the rest of the afternoon.
         | 
         | GPT-3 is closer to a talking toaster than a Parrot.
        
       ___________________________________________________________________
       (page generated 2022-10-07 23:01 UTC)