3 ms·

The problem with this facile view of things is that it seems to be a dead end for scientific theories. What if we just limited the science of birds to explainin

The problem with this facile view of things is that it seems to be a dead end for scientific theories. What if we just limited the science of birds to explaining how limb-flapping could produce levitation? Hmm yes. Birds are kind of like helicopters, it seems. Who’s to say that they are not basically one and the same? Moving on. If you are only interested in the most superficial tests and theories—like the Turing Test—then consider psychology conquered once you’ve tricked a human with your chat bot. Game Over. And what did you learn...?

HTTP418 4y ago
> Birds are kind of like helicopters, it seems. We could easily argue that birds are not a type of helicopter because for helicopter's we have a very specific set of flying properties required. It must have a main propeller for lift and a tail propeller to counter balance the main propeller from spinning the helicopter. If a bird flew with similar mechanism I would argue it was a helicopter. We don't have a 100% accurate gauge for ToM as far as we know. This paper simply uses some of the best known tests for ToM and then states that either LLM can lead to emergent properties or that the current tests for ToM need to be re-thought.
curiousllama 4y ago
> What if we just limited the science of birds to explaining how limb-flapping could produce levitation? Hmm yes. Birds are kind of like helicopters, it seems. Who’s to say that they are not basically one and the same? Moving on. Well said. I'm gonna steal this explanation. Also reminds me of the famous Carbonara quote: "if my grandmother had wheels, then she would be a bike" [1] [1] https://www.youtube.com/watch?v=A-RfHC91Ewc https://www.youtube.com/watch?v=A-RfHC91Ewc
- HTTP418 4y ago
  > "if my grandmother had wheels, then she would be a bike" Well it could be argued that she would be a bike. Its possible to be multiple things at once. If she had 2 wheels and could be ridden by other humans to a destination she might qualify has a bike. She would also continue to be your grandmother.
s0ulphire 4y ago
The question being asked was "what it would mean under this definition of knowing things for a machine learning algorithm to ever know something". Aside from your answer being rude, it's also unhelpful in that it doens't address the question asked and instead relies on reductio ad absurdum to pretend to make a point. If you'd like to take a crack at a helpful answer, perhaps educate us all on what it WOULD take for you to consider a NN to actually "know" something in the same way that we say a human or other sentient animal does.
- avgcorrection 4y ago
  > Aside from your answer being rude, it's also unhelpful in that it doens't address the question asked and instead relies on reductio ad absurdum to pretend to make a point. That is indeed often the kind of answer that a philosophical question deserves. > If you'd like to take a crack at a helpful answer, perhaps educate us all on what it WOULD take for you to consider a NN to actually "know" something in the same way that we say a human or other sentient animal does. How many angels can dance on the head of a pin?
bryan0 4y ago
> The problem with this facile view of things is that it seems to be a dead end for scientific theories. You're overreaching quite a bit here, or I think you're misinterpreting what Parent said. I interpreted what they said as: it seems the difference in how we "know" something vs how an LLM "knows" something might actually be closer than some suspect. this certainly is not an "end of science".
- avgcorrection 4y ago
  Not the end. Just the exact opposite attitude that an inquisitive and humble scientist should have. So unlikely to go anywhere. A “scientist” looks out at his living room. My Roomba and my cat have their own lifes. Who’s to say that they are not in fact the same in kind (but not degreee)? Good luck with that, professor.
shawnz 4y ago
> If you are only interested in the most superficial tests and theories—like the Turing Test—then consider psychology conquered once you’ve tricked a human with your chat bot. What's the counterargument? What's a less superficial test that we can use instead, which conclusively shows that actually human minds aren't just like very sophisticated LLMs? There isn't one -- this is nothing but the same Chinese room problem which we've been discussing for decades. The topmost poster is simply assuming that language models can't possibly understand the same way a human does without relying on any kind of "test" at all, which I think is the real scientific dead end here.
- avgcorrection 4y ago
  > (though in general I think the favored “alignment” frames of the LessWrong community are not even wrong). The Turing Test doesn’t test humans. So you cannot use it to show any properties about humans. Next! > The topmost poster is simply assuming that language models can't possibly understand the same way a human does without relying on any kind of "test" at all, which I think is the real scientific dead end here. Sounds unfalsifiable. So yes.
  - shawnz 4y ago
    If you are actually interested in this problem why not try interpreting what I'm saying a bit more charitably and not waste your time replying with snark?
psb217 4y ago
How is the request that someone provide a clear set of definitions and some empirically falsifiable hypotheses a "dead end for scientific theories"? It seems more like the foundation of the scientific method.
- avgcorrection 4y ago
  > and some empirically falsifiable hypotheses Where?

HTTP418 4y ago

> Birds are kind of like helicopters, it seems. We could easily argue that birds are not a type of helicopter because for helicopter's we have a very specific set of flying properties required. It must have a main propeller for lift and a tail propeller to counter balance the main propeller from spinning the helicopter. If a bird flew with similar mechanism I would argue it was a helicopter. We don't have a 100% accurate gauge for ToM as far as we know. This paper simply uses some of the best known tests for ToM and then states that either LLM can lead to emergent properties or that the current tests for ToM need to be re-thought.

curiousllama 4y ago

> What if we just limited the science of birds to explaining how limb-flapping could produce levitation? Hmm yes. Birds are kind of like helicopters, it seems. Who’s to say that they are not basically one and the same? Moving on. Well said. I'm gonna steal this explanation. Also reminds me of the famous Carbonara quote: "if my grandmother had wheels, then she would be a bike" [1] [1] https://www.youtube.com/watch?v=A-RfHC91Ewc https://www.youtube.com/watch?v=A-RfHC91Ewc

HTTP418 4y ago
> "if my grandmother had wheels, then she would be a bike" Well it could be argued that she would be a bike. Its possible to be multiple things at once. If she had 2 wheels and could be ridden by other humans to a destination she might qualify has a bike. She would also continue to be your grandmother.

s0ulphire 4y ago

The question being asked was "what it would mean under this definition of knowing things for a machine learning algorithm to ever know something". Aside from your answer being rude, it's also unhelpful in that it doens't address the question asked and instead relies on reductio ad absurdum to pretend to make a point. If you'd like to take a crack at a helpful answer, perhaps educate us all on what it WOULD take for you to consider a NN to actually "know" something in the same way that we say a human or other sentient animal does.

avgcorrection 4y ago
> Aside from your answer being rude, it's also unhelpful in that it doens't address the question asked and instead relies on reductio ad absurdum to pretend to make a point. That is indeed often the kind of answer that a philosophical question deserves. > If you'd like to take a crack at a helpful answer, perhaps educate us all on what it WOULD take for you to consider a NN to actually "know" something in the same way that we say a human or other sentient animal does. How many angels can dance on the head of a pin?

bryan0 4y ago

> The problem with this facile view of things is that it seems to be a dead end for scientific theories. You're overreaching quite a bit here, or I think you're misinterpreting what Parent said. I interpreted what they said as: it seems the difference in how we "know" something vs how an LLM "knows" something might actually be closer than some suspect. this certainly is not an "end of science".

avgcorrection 4y ago
Not the end. Just the exact opposite attitude that an inquisitive and humble scientist should have. So unlikely to go anywhere. A “scientist” looks out at his living room. My Roomba and my cat have their own lifes. Who’s to say that they are not in fact the same in kind (but not degreee)? Good luck with that, professor.

shawnz 4y ago

> If you are only interested in the most superficial tests and theories—like the Turing Test—then consider psychology conquered once you’ve tricked a human with your chat bot. What's the counterargument? What's a less superficial test that we can use instead, which conclusively shows that actually human minds aren't just like very sophisticated LLMs? There isn't one -- this is nothing but the same Chinese room problem which we've been discussing for decades. The topmost poster is simply assuming that language models can't possibly understand the same way a human does without relying on any kind of "test" at all, which I think is the real scientific dead end here.

avgcorrection 4y ago
> (though in general I think the favored “alignment” frames of the LessWrong community are not even wrong). The Turing Test doesn’t test humans. So you cannot use it to show any properties about humans. Next! > The topmost poster is simply assuming that language models can't possibly understand the same way a human does without relying on any kind of "test" at all, which I think is the real scientific dead end here. Sounds unfalsifiable. So yes.
- shawnz 4y ago
  If you are actually interested in this problem why not try interpreting what I'm saying a bit more charitably and not waste your time replying with snark?

psb217 4y ago

How is the request that someone provide a clear set of definitions and some empirically falsifiable hypotheses a "dead end for scientific theories"? It seems more like the foundation of the scientific method.

avgcorrection 4y ago
> and some empirically falsifiable hypotheses Where?