5 ms·

Too many googleable questions (like who was the president). Too little 'understanding' type questions like 'Why don't animals have three legs?' In addition to

Too many googleable questions (like who was the president). Too little 'understanding' type questions like 'Why don't animals have three legs?' In addition to nonsense questions, I think it would be pretty easy to knock it over with some deeper questions about things they were already talking about. Like asking 'Why don't chairs with 3 legs fall over then?'

Micoloth 6y ago
> Too many googleable questions To be fair (as one who hasn't paid too much attention to these developements in the last few months/years), the sheer amount of individual facts that are apparently encoded in a fixed nn architecture is what amazes me the most here? I get that it has 175B variables, but it's not like you can used them in a structured way to store say a knowledge base...
- rainbow-road 6y ago
  I see this "175B parameters" figure tossed around a lot: what does that mean? Does the model literally consist of 175 billion 32-bit floats? (Sorry for the very basic question.)
  - qayxc 6y ago
    You would be correct. It's ~650GiB of data. The entire English version of Wikipedia is about 20GiB and the Encyclopædia Britannica is only about 300MiB (measured in ASCII characters) [1]. [1] https://en.wikipedia.org/wiki/Wikipedia:Size_comparisons https://en.wikipedia.org/wiki/Wikipedia:Size_comparisons
    - pradn 6y ago
      The English Wikipedia text (no images) is about ~20 GB compressed, but ~60 GB uncompressed.
      - qayxc 6y ago
        I only use the sources available to me at the time. Another value published by Wikipedia is 30GiB [1] as of 2020, which includes punctuation and markup. I explicitly put the measurement unit as ASCII characters. If you have a better source for your size (remember: ASCII characters for article words only, no markup), feel free to post it. [1] https://en.wikipedia.org/wiki/Wikipedia:Size_in_volumes https://en.wikipedia.org/wiki/Wikipedia:Size_in_volumes
andybak 6y ago
That's the part of the Turing Test I've never understood - it seems very dependent on the skill and intelligence of the human tester. Did Turing talk about this aspect? I seem to remember there was supposed to be a panel or committee rather than a single individual?
- nurettin 6y ago
  Turing explains that the machine could not give an "appropriately meaningful" answer to whatever is said in it's presence, which the "dullest of men" could do. So the basic quality he seeks in a machine that could pass the so called turing test is a machine that produces meaning. That meaning comes from any mind that is intact or coherent. So unless we take the leap from making category deciders and inference machines to actually building minds, we won't get anywhere near his definition of a machine that is truly intelligent.
  - zwischenzug 6y ago
    Is this a real comment, or produced by an AI? I couldn't make sense of it. This is a genuine question.
    - oldsj 6y ago
      I think the cat’s out of the bag now and we really won’t know if we’re reading or responding to an actual human anymore.
      - gear54rus 6y ago
        Reading comments on the internet has nothing to do with the test tho. You have to be able to query the agent.
        andybak 6y ago
        > You have to be able to query the agent. That's what the "reply" button is for.
paraschopra 6y ago
GPT-3 actually knows: https://imgur.com/a/DHvb8Iu https://imgur.com/a/DHvb8Iu

Micoloth 6y ago

> Too many googleable questions To be fair (as one who hasn't paid too much attention to these developements in the last few months/years), the sheer amount of individual facts that are apparently encoded in a fixed nn architecture is what amazes me the most here? I get that it has 175B variables, but it's not like you can used them in a structured way to store say a knowledge base...

rainbow-road 6y ago
I see this "175B parameters" figure tossed around a lot: what does that mean? Does the model literally consist of 175 billion 32-bit floats? (Sorry for the very basic question.)
- qayxc 6y ago
  You would be correct. It's ~650GiB of data. The entire English version of Wikipedia is about 20GiB and the Encyclopædia Britannica is only about 300MiB (measured in ASCII characters) [1]. [1] https://en.wikipedia.org/wiki/Wikipedia:Size_comparisons https://en.wikipedia.org/wiki/Wikipedia:Size_comparisons
  - pradn 6y ago
    The English Wikipedia text (no images) is about ~20 GB compressed, but ~60 GB uncompressed.
    - qayxc 6y ago
      I only use the sources available to me at the time. Another value published by Wikipedia is 30GiB [1] as of 2020, which includes punctuation and markup. I explicitly put the measurement unit as ASCII characters. If you have a better source for your size (remember: ASCII characters for article words only, no markup), feel free to post it. [1] https://en.wikipedia.org/wiki/Wikipedia:Size_in_volumes https://en.wikipedia.org/wiki/Wikipedia:Size_in_volumes

andybak 6y ago

That's the part of the Turing Test I've never understood - it seems very dependent on the skill and intelligence of the human tester. Did Turing talk about this aspect? I seem to remember there was supposed to be a panel or committee rather than a single individual?

nurettin 6y ago
Turing explains that the machine could not give an "appropriately meaningful" answer to whatever is said in it's presence, which the "dullest of men" could do. So the basic quality he seeks in a machine that could pass the so called turing test is a machine that produces meaning. That meaning comes from any mind that is intact or coherent. So unless we take the leap from making category deciders and inference machines to actually building minds, we won't get anywhere near his definition of a machine that is truly intelligent.
- zwischenzug 6y ago
  Is this a real comment, or produced by an AI? I couldn't make sense of it. This is a genuine question.
  - oldsj 6y ago
    I think the cat’s out of the bag now and we really won’t know if we’re reading or responding to an actual human anymore.
    - gear54rus 6y ago
      Reading comments on the internet has nothing to do with the test tho. You have to be able to query the agent.
      - andybak 6y ago
        > You have to be able to query the agent. That's what the "reply" button is for.

paraschopra 6y ago

GPT-3 actually knows: https://imgur.com/a/DHvb8Iu https://imgur.com/a/DHvb8Iu