7 ms·

AI Is Not a Horse

hprotagonist 4y ago
Hmm. my metaphor for AI is as a statistical prediction machine that produces maximally likely output for a given input, conditioned on some historical data.
- viscanti 4y ago
  Well maybe for DNs but there can be deterministic systems that are learned through RL that aren't stochastic. But it's a good enough metaphor for a big part of what people talk about when they say AI.
- freitzkriesler 4y ago
  I've been using "AI is algorithmic slight of hand."
  - jfghi 4y ago
    I’d say FGPM for fairly good pattern matching would be a nice descriptor.
  - zmgsabst 4y ago
    In what sense is AI “sleight of hand”, but your intelligence isn’t?
    - freitzkriesler 4y ago
      Real human intelligence comes from a place where we don't understand yet. I recall experiments being done where they could identify where a "thought" started but now how it initiated. Have to find the research now. Whereas AI are just algorithms which read data and process it into schemas IE we know the beginning, middle, and end and is nothing more than a facsimile of what we know as real bonafide intelligence from a human.
- visarga 4y ago
  Your metaphor fails to indicate the dependence of AI on data. AI can't appear in a void. There must be an environment producing its data, then AI development is conditioned on the environment. We forget about that when we just focus on the learned model. But without all the external support structure there is no AI. Let's take a simple case - AlphaZero. Its environment was just a Go board, and rules simple. Knowing who won the game easy to test at the end of the game. So this led to a massive number of self-play games. In a few days AZ beat humans, and it was our own game, with 2 millennia head start. This is how much the environment and the external support structure influences AI development. "Maximally likely output for a given input" depends on it.
  - mcswell 4y ago
    "Your metaphor fails to indicate the dependence of AI on data." I'm not sure how you conclude that; the post you're responding to says exactly "...conditioned on some historical data." No historical data, no conditioning, and therefore no AI, just random bits. And fwiw, you could say much the same about human beings. There must be an environment producing our oxygen, water, food. Indeed, it's not clear what a single human raised in isolation would be like, even assuming the oxygen, water, food. Not very intelligent seeming, I would guess; perhaps not even with much of a language. Fortunately, no one has performed the experiment, but we've come close: deaf people raised by non-deaf parents in a non-deaf community appear to have a sparse sort of sign language. Only when they're put together with other deaf people does a more typical, fluent sign language arise, with a full grammar. (This we know from the deaf people in Nicaragua.)
fragmede 4y ago
Desktop computers have been around since there 1960's, and Apple's first computer was released in 1974. I think people understand what computer programs are without some clumbsy or awkward metaphor. ChatGPT isn't sentient, it's a large language model (LLM) that performs better than a Markov Chain but the underlying principal isn't so far off, but it's still a computer program. One that you can anthroponorphize readily, but it's still just a computer program, albiet a modern, and quite impressive one. It's a tool like a hammer
- visarga 4y ago
  > but it's still just a computer program Maybe you are not impressed by the answer to "how does it work?", but we need to think about "what does it learn?". What data, what skills, what limitations. It's not just any data, it is our whole culture as a dataset, does it have some kind of special properties?
  - YeGoblynQueenne 4y ago
    You're absolutely right that we should try to understand LLMs. That's the point of science, after all, to understand stuff. But the truth is that very few people are actually trying to understand LLMs. Rather, what the vast majority do, is that they poke them unsystematically and then ooh and aah about all the things they can do that earlier language models couldn't do. Careful analysis? Thorough, systematic experimentation? Theoretical results???? I have yet to see any of that and of course it's a joke to talk about theoretical results when it comes to machine learning research, these days. It's as if we don't have any computer science, any theories of information, complexity, computation, learnability, etc, that we could muster to try and explain how a complex computer program works. So are people really trying to understand or are they just happy to ride the hype train? Because I obviously think it's the latter.
- rhn_mk1 4y ago
  Who's to say that a brain is not a kind of a computer and that all humans are a sort of a computer program? Going from "computer program" to "not sentient" is a leap of logic I'd like to hear more about.
lowbloodsugar 4y ago
None of this “AI” is actually AI. When we do finally meet one, we’ll probably be its gut bacteria.
- Animats 4y ago
  AIs, once they become better at business than humans, will be in charge. That is implicit in the concept of corporate capitalism.
  - Tijdreiziger 4y ago
    The shareholders will be in charge. Of course, it’s possible that trading firms will choose to outsource the job to AIs.
GaggiX 4y ago
I don't think we need a metaphor to understand the relationship between humans and AIs; AIs are tools used by humans like many others.
- programmarchy 4y ago
  Ah yes, and the Internet is just a series of tubes.
  - GaggiX 4y ago
    I don't know what your strange definition of the internet has to do with my comment, but just to stay on topic, the internet is also a tool we used, to communicate or inform ourselves, we are talking about the relationship between thing X and humans, not its definition.
    - programmarchy 4y ago
      It’s a metaphor (a bad one) made famous by a US Senator. I used it to invoke the importance of metaphor. We absolutely need metaphors (good ones) to explain technology to not only the public, but also to the people who legislate. The Internet is not just a tool in the same way that a hammer or a dishwasher is a tool because it has had a totally transformative effect on humanity. It’s on the level of fire, gun powder, or the printing press. And there are inklings that the same argument could be made for AI. So your comment is just completely missing the point. “The medium is the message.”
      - GaggiX 4y ago
        One should use good (hopefully) metaphors to explain technology as you said, this is done all the time, but here we want to define the relationship between humans and thing X, and I would say there is no difference in the relationship with AIs as with all other tools, I would not treat a hammer differently from a printing press because it has less transformative effect on humanity.
thom 4y ago
Oh, I think on a long enough timeline the master/slave metaphor will prove apt, possibly twice.
protocolture 4y ago
AI is part of a Centaur per Kasparov
User23 4y ago
I’m curious, does the author actually have any horses?
- booleandilemma 4y ago
  Are you gatekeeping a metaphor?
  - User23 4y ago
    Don’t be rude please.
- georgestrakhov 4y ago
  As a matter of fact, I don't. But my daughter does horse riding and volunteers at the stables and the whole thing is quite beautiful. Though it smells. Something to add to the metaphor maybe.
tommykins 4y ago
Nick Cave and HackerNews intersection is not something I saw coming.
grantcas 4y ago
[dead]

Hmm. my metaphor for AI is as a statistical prediction machine that produces maximally likely output for a given input, conditioned on some historical data.

viscanti 4y ago
Well maybe for DNs but there can be deterministic systems that are learned through RL that aren't stochastic. But it's a good enough metaphor for a big part of what people talk about when they say AI.
freitzkriesler 4y ago
I've been using "AI is algorithmic slight of hand."
- jfghi 4y ago
  I’d say FGPM for fairly good pattern matching would be a nice descriptor.
- zmgsabst 4y ago
  In what sense is AI “sleight of hand”, but your intelligence isn’t?
  - freitzkriesler 4y ago
    Real human intelligence comes from a place where we don't understand yet. I recall experiments being done where they could identify where a "thought" started but now how it initiated. Have to find the research now. Whereas AI are just algorithms which read data and process it into schemas IE we know the beginning, middle, and end and is nothing more than a facsimile of what we know as real bonafide intelligence from a human.
visarga 4y ago
Your metaphor fails to indicate the dependence of AI on data. AI can't appear in a void. There must be an environment producing its data, then AI development is conditioned on the environment. We forget about that when we just focus on the learned model. But without all the external support structure there is no AI. Let's take a simple case - AlphaZero. Its environment was just a Go board, and rules simple. Knowing who won the game easy to test at the end of the game. So this led to a massive number of self-play games. In a few days AZ beat humans, and it was our own game, with 2 millennia head start. This is how much the environment and the external support structure influences AI development. "Maximally likely output for a given input" depends on it.
- mcswell 4y ago
  "Your metaphor fails to indicate the dependence of AI on data." I'm not sure how you conclude that; the post you're responding to says exactly "...conditioned on some historical data." No historical data, no conditioning, and therefore no AI, just random bits. And fwiw, you could say much the same about human beings. There must be an environment producing our oxygen, water, food. Indeed, it's not clear what a single human raised in isolation would be like, even assuming the oxygen, water, food. Not very intelligent seeming, I would guess; perhaps not even with much of a language. Fortunately, no one has performed the experiment, but we've come close: deaf people raised by non-deaf parents in a non-deaf community appear to have a sparse sort of sign language. Only when they're put together with other deaf people does a more typical, fluent sign language arise, with a full grammar. (This we know from the deaf people in Nicaragua.)

fragmede 4y ago

Desktop computers have been around since there 1960's, and Apple's first computer was released in 1974. I think people understand what computer programs are without some clumbsy or awkward metaphor. ChatGPT isn't sentient, it's a large language model (LLM) that performs better than a Markov Chain but the underlying principal isn't so far off, but it's still a computer program. One that you can anthroponorphize readily, but it's still just a computer program, albiet a modern, and quite impressive one. It's a tool like a hammer

visarga 4y ago
> but it's still just a computer program Maybe you are not impressed by the answer to "how does it work?", but we need to think about "what does it learn?". What data, what skills, what limitations. It's not just any data, it is our whole culture as a dataset, does it have some kind of special properties?
- YeGoblynQueenne 4y ago
  You're absolutely right that we should try to understand LLMs. That's the point of science, after all, to understand stuff. But the truth is that very few people are actually trying to understand LLMs. Rather, what the vast majority do, is that they poke them unsystematically and then ooh and aah about all the things they can do that earlier language models couldn't do. Careful analysis? Thorough, systematic experimentation? Theoretical results???? I have yet to see any of that and of course it's a joke to talk about theoretical results when it comes to machine learning research, these days. It's as if we don't have any computer science, any theories of information, complexity, computation, learnability, etc, that we could muster to try and explain how a complex computer program works. So are people really trying to understand or are they just happy to ride the hype train? Because I obviously think it's the latter.
rhn_mk1 4y ago
Who's to say that a brain is not a kind of a computer and that all humans are a sort of a computer program? Going from "computer program" to "not sentient" is a leap of logic I'd like to hear more about.

lowbloodsugar 4y ago

None of this “AI” is actually AI. When we do finally meet one, we’ll probably be its gut bacteria.

Animats 4y ago
AIs, once they become better at business than humans, will be in charge. That is implicit in the concept of corporate capitalism.
- Tijdreiziger 4y ago
  The shareholders will be in charge. Of course, it’s possible that trading firms will choose to outsource the job to AIs.

GaggiX 4y ago

I don't think we need a metaphor to understand the relationship between humans and AIs; AIs are tools used by humans like many others.

programmarchy 4y ago
Ah yes, and the Internet is just a series of tubes.
- GaggiX 4y ago
  I don't know what your strange definition of the internet has to do with my comment, but just to stay on topic, the internet is also a tool we used, to communicate or inform ourselves, we are talking about the relationship between thing X and humans, not its definition.
  - programmarchy 4y ago
    It’s a metaphor (a bad one) made famous by a US Senator. I used it to invoke the importance of metaphor. We absolutely need metaphors (good ones) to explain technology to not only the public, but also to the people who legislate. The Internet is not just a tool in the same way that a hammer or a dishwasher is a tool because it has had a totally transformative effect on humanity. It’s on the level of fire, gun powder, or the printing press. And there are inklings that the same argument could be made for AI. So your comment is just completely missing the point. “The medium is the message.”
    - GaggiX 4y ago
      One should use good (hopefully) metaphors to explain technology as you said, this is done all the time, but here we want to define the relationship between humans and thing X, and I would say there is no difference in the relationship with AIs as with all other tools, I would not treat a hammer differently from a printing press because it has less transformative effect on humanity.

thom 4y ago

Oh, I think on a long enough timeline the master/slave metaphor will prove apt, possibly twice.

protocolture 4y ago

AI is part of a Centaur per Kasparov

User23 4y ago

I’m curious, does the author actually have any horses?

booleandilemma 4y ago
Are you gatekeeping a metaphor?
- User23 4y ago
  Don’t be rude please.
georgestrakhov 4y ago
As a matter of fact, I don't. But my daughter does horse riding and volunteers at the stables and the whole thing is quite beautiful. Though it smells. Something to add to the metaphor maybe.

tommykins 4y ago

Nick Cave and HackerNews intersection is not something I saw coming.

grantcas 4y ago

[dead]