• 0 Posts
  • 31 Comments
Joined 2 years ago
cake
Cake day: August 2nd, 2023

help-circle

  • Ikea really managed to pull off a magnificent marketing stunt.

    They have the same furniture quality that you can get off of Amazon, Wayfair, or Walmart but you have to go get it out of their warehouse and deal with the logistics of getting it to your house. But they hand out some meatballs and give everything funky Swedish names; so people get the impression that it’s a fancy European experience.

    China is the single largest manufacturer of Ikea products. I found that when I went directly to the source, I could get the same item cheaper or a better item at the same price. That deal is likely to die with the new tariff regime but that same regime will have similar impacts on Ikea.



  • nednobbins@lemm.eetoScience Memes@mander.xyzPercentages
    link
    fedilink
    English
    arrow-up
    2
    ·
    7 months ago

    Yes. I really think that the translation part is one of the hardest.

    As a brief aside, I want to note that this conversation is happening in one of the languages we’re discussing and that could influence any conclusion we come to. I’m also going to suggest that we ignore Gödel for now

    There are many people who are good at math. There are even a lot of people who are reasonably good at grinding through the mechanics of math. That doesn’t solve any of the problems you described above.

    Statistics are a great example of this. Early statistics classics are mostly about the mechanics; here’s how you calculate the mean, standard deviation, confidence intervals, etc. 2 types of students generally come out of that class; math students who will forget all of that because they’re going to learn the “real” versions (eg they go through a huge number of proofs that involve calculus and linear algebra), and students who will forget all of that because the whole thing sounds like gibberish.

    We teach natural languages the same way but we go much farther. Students learn vocabulary and grammar rules but they’re also expected to learn how to use them correctly. We had students current events articles and ask them to analyze them. We ask students to practice many writing methods including fiction and expository writing.

    When I talk to my own kids about statistics I never write any formulas. I ask questions like, “What do you think ‘mean’ means?”, “If I have a bunch of <example item> does ‘mean’ describe it well?”, “What happens if I add an <example item> with <huge outlier>? Do you still think it’s a good description?” “How would you describe it better?”

    If I ever had to design an introductory statistics course it would contain very little “math”. Classes, homework, projects, and tests would consist of questions like; “Here’s some data and an interpretation, are they lying? Why or why not?” “Here’s a (simple) data scenario. Tell me what’s going on.” “Here’s some (simple) data. Produce a correct and faithful summary. Now produce a correct but misleading summary. Describe what you did and the effect.” “Here’s a conclusion. Provide sample data that most likely fits the conclusion.” “Change one word in the sentence, ‘Increase your chances by 80% means that there is now an 80% chance.’ to make it a true statement.”


  • nednobbins@lemm.eetoScience Memes@mander.xyzPercentages
    link
    fedilink
    English
    arrow-up
    16
    arrow-down
    1
    ·
    7 months ago

    Why not both?

    I’ve always thought of math as a language and I talk to my kids about it that way too. Math is an other way to describe the world.

    It’s very different from spoken languages and translating between the two needs to be learned and practiced.

    Our math education doesn’t include enough word problems and it should be bi-directional. In addition to teaching students how to write equations based of sentences we should teach them how to describe what’s going on in an equation.








  • I get the feeling of discomfort but it’s basically the same feeling we get when someone breaks a pencil

    There is no evidence that a mosquito is capable of feeling the kind of despair or horror that a human would feel in a similar situation. It’s unlikely that mosquitos can form emotions at all.

    At the same time, a huge portion of human-animal interactions involve the human controlling the animal in ways that they animal can’t even comprehend. A dog has no idea you’re doing operant conditioning to change their behavior. Pigs have no idea they’re being fed just so they and their children can be eaten.

    The only way to avoid this kind of thing is to turn off your big human brain and go back to ape tier. We might need to go farther down the tier list than that though https://en.m.wikipedia.org/wiki/Gombe_Chimpanzee_War









  • Covering the second half:

    I hadn’t heard of Elsagate and had to look it up. How does AI factor into that? As near as I can tell Elsagate started with some random guy making disturbing videos and mislabeling them as child-friendly.

    I’m a good bit older than you so my nostalgia doesn’t take me lead me to any of the title you mentioned. For the most part it’s stories that aren’t covered by anyone’s IP. My childhood had a lot of folk tales recited from memory. Those stories were fairly common but there would be regional variation and most tellers would put their own twist on the stories (for example, when my Aunt told the story of the Seven Kids she would do a particular squeaky voice when she got to the part where the wolf swallows the chalk (in her version it was always chalk). That’s actually quite close to how LLMs work. She heard various versions of that story throughout her life, then she repeats it with some other bits that she incorporated from the rest of her life. I do the same thing when I retell the story to my children. It’s basically the same story my Aunt told but I translate it into English and add some modern slang.

    What would stop an AI from writing Scar into the Lion King? If you told an LLM to, “Write Hamlet but have all the royal family be Lions,” it’s likely you’d get some evil lion version of Claudius.

    There were a lot of homosexual coded villains in older media. There were also a lot of films where all the black people were bad guys, all the Asian people were goofy servants and all the women were housewives or prizes. The general consensus today is that those choices were horribly discriminatory. If AI manages to avoid that sort of behavior it would be a good thing.

    The flip side is also that artists can just as easily slip hateful material into otherwise reasonable art. Human history is full of unethical choices. Even if the AI itself doesn’t have ethics the people using it can be held to the same ethical standards as the users of any other tool or medium.