3 ms·

Eventually it will be common knowledge that asking a text compressor trained on all available text to design a new technology doesn't work.

alchemist1e9 4y ago
I don’t believe that is true. Most technologies are incremental improvements on recombinations of existing tools and techniques. It seem likely to me that LLMs ability to map associations between concepts can result in humans using it to help invent new technology. Have you personally used GPT-4 much?
- inimino 4y ago
  I try to base my judgments of what LLMs can and can't do primarily on my study and research in related fields. I haven't been surprised by the capabilities of any LLM yet, including GPT-4.
  - alchemist1e9 4y ago
    Are you kidding me? Pretty much everyone is surprised, including the creators, how are you not?
    - inimino 4y ago
      Is that a serious question? Studying a field for years should make outcomes in that field less surprising, otherwise what have you been doing? The creators were surprised in the sense of "we got here sooner than expected" but not "we didn't think this would work". Otherwise they wouldn't have been working on it. And there is nothing new in LLMs in years, it's just increasing fidelity by massively increased scale. To be honest, I've been more surprised by the incompetence of people in evaluating these systems, including journalists, programmers, and others who should be in a position to know better.
      - alchemist1e9 4y ago
        > The creators were surprised in the sense of "we got here sooner than expected" but not "we didn't think this would work". Otherwise they wouldn't have been working on it. And there is nothing new in LLMs in years, it's just increasing fidelity by massively increased scale. This is categorically false. There are papers being published on all the surprising emergent behavior being observed.
        inimino 4y ago
        Emergent, as in, "we cannot explain how this works," yes. That is nothing new in the field of ML or to anyone who has been paying attention.
        alchemist1e9 4y ago
        https://youtu.be/StLtMcsbQes https://youtu.be/StLtMcsbQes I think you have your head in the sand and haven’t been paying attention. The scaling laws are not expected. The capabilities of GPT-3.5 are beyond what even those deeply involved had expected. I also think the progress is likely going exponential at this point. Multi agent and recursive prompting are coming soon. This is really not ML at all. I have extensive traditional ML knowledge and background. I know in detail the typical model suspects on a Kaggle board. LLMs are totally new and surprising relative to my many decades working with ML and traditional NLP.
        inimino 4y ago
        That's a good talk. I'm paying attention. I think "scale is all you need" is wrong even when it's right. We have a responsibility to not allow the capabilities to outstrip our ability to understand and control. If we don't do our job that will be the real "bitter lesson." However, ultimately it's a text predictor driven by a PRNG and I stand by my statement; I think the systems are obviously impressive but the unrealistic expectations people have and the anthropomorphization and projection I'm seeing is even more impressive. Let me know when it starts synthesizing new science or math. By then we're in trouble.

alchemist1e9 4y ago

I don’t believe that is true. Most technologies are incremental improvements on recombinations of existing tools and techniques. It seem likely to me that LLMs ability to map associations between concepts can result in humans using it to help invent new technology. Have you personally used GPT-4 much?

inimino 4y ago
I try to base my judgments of what LLMs can and can't do primarily on my study and research in related fields. I haven't been surprised by the capabilities of any LLM yet, including GPT-4.
- alchemist1e9 4y ago
  Are you kidding me? Pretty much everyone is surprised, including the creators, how are you not?
  - inimino 4y ago
    Is that a serious question? Studying a field for years should make outcomes in that field less surprising, otherwise what have you been doing? The creators were surprised in the sense of "we got here sooner than expected" but not "we didn't think this would work". Otherwise they wouldn't have been working on it. And there is nothing new in LLMs in years, it's just increasing fidelity by massively increased scale. To be honest, I've been more surprised by the incompetence of people in evaluating these systems, including journalists, programmers, and others who should be in a position to know better.
    - alchemist1e9 4y ago
      > The creators were surprised in the sense of "we got here sooner than expected" but not "we didn't think this would work". Otherwise they wouldn't have been working on it. And there is nothing new in LLMs in years, it's just increasing fidelity by massively increased scale. This is categorically false. There are papers being published on all the surprising emergent behavior being observed.
      - inimino 4y ago
        Emergent, as in, "we cannot explain how this works," yes. That is nothing new in the field of ML or to anyone who has been paying attention.
        alchemist1e9 4y ago
        https://youtu.be/StLtMcsbQes https://youtu.be/StLtMcsbQes I think you have your head in the sand and haven’t been paying attention. The scaling laws are not expected. The capabilities of GPT-3.5 are beyond what even those deeply involved had expected. I also think the progress is likely going exponential at this point. Multi agent and recursive prompting are coming soon. This is really not ML at all. I have extensive traditional ML knowledge and background. I know in detail the typical model suspects on a Kaggle board. LLMs are totally new and surprising relative to my many decades working with ML and traditional NLP.
        inimino 4y ago
        That's a good talk. I'm paying attention. I think "scale is all you need" is wrong even when it's right. We have a responsibility to not allow the capabilities to outstrip our ability to understand and control. If we don't do our job that will be the real "bitter lesson." However, ultimately it's a text predictor driven by a PRNG and I stand by my statement; I think the systems are obviously impressive but the unrealistic expectations people have and the anthropomorphization and projection I'm seeing is even more impressive. Let me know when it starts synthesizing new science or math. By then we're in trouble.