AI is boring

Why did no one tell me that AI is so boring? I thought I was going to learn about ChatGPT and machine learning but instead it's just an algorithms class with every problem being a state tree or state graph.

Mike Stoklasa's Worst Fan Shirt $21.68

A Conspiracy Theorist Is Talking Shirt $21.68

Mike Stoklasa's Worst Fan Shirt $21.68

  1. 7 months ago
    Anonymous

    anon that's what machine learning is, you are learning about ML

    • 7 months ago
      Anonymous

      Closet thing to machine learning I've done in that class is use simulated annealing to generate policies for a problem. We don't even touch q-tables in my AI class and we spent like 4 weeks on breath first search since the concept of a state tree filtered the class and dumbed everything down.

      • 7 months ago
        Anonymous

        That's because Machine Learning is a subset of AI.

  2. 7 months ago
    Anonymous

    if you want to understand how these models work then you need to know this. if you don’t care then you could have just watched some yt video breaking it down, no need for the course

    • 7 months ago
      Anonymous

      That's why theres prerequisites such as data structures and algorithms so that the class doesn't have to spend a couple weeks on what a state tree is or how a* is different from uniform cost. AI college courses are just algorithms with a different skin thrown on it.

      • 7 months ago
        Anonymous

        I mean ai covers more classical techniques besides machine learning right, that’s the point of the course. makes sense it covers these topics

        • 7 months ago
          Anonymous

          First, you have to understand: AI is just a fancy phrase for "stuff we don't have a good time complexity algorithm, or problems where the structure isn't well enough understood to yield a closed form solution". BFS is AI, you only use techniques like that if no simple solution exists.

          They teach the simpler methods because morons will just apply whatever the hot ticket AI algorithm is without any thought that there might be a more effective technique. Just because you can do with an LLM doesn't mean you should.

          I suppose you both have a point. I wish there communication on what AI but that's partly my fault for expecting something different.

          [...]
          Is BFS AI? I was under the impression that it's just a complete search algorithm on a graph.

          You can modify BFS to act as an AI agent that can solve problems thay you can map to a state graph or state tree which makes it a form of AI. AI isn't as glamorous as it seems.

          • 7 months ago
            Anonymous

            Nah, it's shit for AI. Technically you want to modify DFS so it runs like BFS, so you get the memory benefit of DFS, but with the completeness of BFS. This in and of itself isnt' AI, though. AI is about phrasing the problem in terms of agents making decisions.
            [...]
            You have it exactly backwards. You reduce the AI problem to a search problem, not the other way 'round.

            I'm still a bit confused, BFS just returns the nodes in order. If you're adding a heuristic to it, doesn't it become A*?

  3. 7 months ago
    Anonymous

    First, you have to understand: AI is just a fancy phrase for "stuff we don't have a good time complexity algorithm, or problems where the structure isn't well enough understood to yield a closed form solution". BFS is AI, you only use techniques like that if no simple solution exists.

    They teach the simpler methods because morons will just apply whatever the hot ticket AI algorithm is without any thought that there might be a more effective technique. Just because you can do with an LLM doesn't mean you should.

  4. 7 months ago
    Anonymous

    What did you expect to do in that class?

  5. 7 months ago
    Anonymous

    [...]

    Is BFS AI? I was under the impression that it's just a complete search algorithm on a graph.

    • 7 months ago
      Anonymous

      Found the second year CS major.

    • 7 months ago
      Anonymous

      Nah, it's shit for AI. Technically you want to modify DFS so it runs like BFS, so you get the memory benefit of DFS, but with the completeness of BFS. This in and of itself isnt' AI, though. AI is about phrasing the problem in terms of agents making decisions.

      [...]
      I suppose you both have a point. I wish there communication on what AI but that's partly my fault for expecting something different.
      [...]
      You can modify BFS to act as an AI agent that can solve problems thay you can map to a state graph or state tree which makes it a form of AI. AI isn't as glamorous as it seems.

      You have it exactly backwards. You reduce the AI problem to a search problem, not the other way 'round.

    • 7 months ago
      Anonymous

      nah CS midwits just love to call every algo involving a graph and traversing it a "BFS problem"

  6. 7 months ago
    Anonymous

    >beta ai student:
    >builds ai libraries
    >chad indie developer:
    >"downloads" that library and simply tells the ai to search for naked cartoons

    which one was smarter?

    • 7 months ago
      Anonymous

      The one that finds 10000 morons to pay for it. Could be anyone of them.

  7. 7 months ago
    Anonymous

    I actually thought it was interesting until all we were doing is deep learning. The classical stuff like kernel methods and bog standard linear classifiers were very fun. But then we started with the black box magic.

  8. 7 months ago
    Anonymous

    >Machine Learning
    >shit don't fit
    >change parameters
    >shit now is overfitting when new data comes in
    >change parameters
    >shit don't fit
    >change parameters
    >shit now is overfitting when new data comes in
    >change parameters
    >shit don't fit
    >change parameters
    >shit now is overfitting when new data comes in
    >change parameters
    >shit don't fit
    >change parameters
    >shit now is overfitting when new data comes in
    >change parameters
    >shit don't fit
    >change parameters
    >shit now is overfitting when new data comes in
    >change parameters

  9. 7 months ago
    Anonymous

    yo

  10. 7 months ago
    Anonymous

    Because you can virtually represent everything as some sort of geometrical representation (including logic and sets), see Topos theory.

  11. 7 months ago
    Anonymous

    I'm in a AI class too like OP and it's slow as shit. But they don't really explain how to code these algo's well. Any resources for this?
    Our last assignment was to code an opposing player (nim) using A* but i couldn't get it quite right.

    • 7 months ago
      Anonymous

      They have pseudo code and actual code for AI algorithms here.
      https://github.com/aimacode

  12. 7 months ago
    Anonymous

    that's because those are the easier things in ML
    you're getting filtered already?

  13. 7 months ago
    Anonymous

    What is that? It doesn't make sense to me,

Your email address will not be published. Required fields are marked *