• 0 Posts
  • 54 Comments
Joined 1 year ago
cake
Cake day: June 10th, 2023

help-circle





  • One of the big advantages of a victorinox is that they’re designed to be essentially maintenance free. As far as I can tell, the intention is that if you leave it in a bag, drawer, car, or just lose it under the couch for a decade, it will be ready to perform when you need it.

    Another great benefit is that you can play around with different maintenance routines and find a system that works for you without worrying about corrosion or excessive wear. Try different oils, try it dry, see how it responds.

    Clean it with water, compressed air, alcohol, or whatever else you feel like trying. Keep in mind that naturally derived oils will go rancid over time and if you’re too thick, it’ll go sticky.

    A similar design philosophy is used with the blade, they are super easy to resharpen. It’s a great blade to learn how to repair and sharpen. It also doesn’t require oiling, but nothing is stopping you from trying it. Just stick to something food-grade so you can use it worry-free on meal prep if you have to.

    Lastly, the most important thing you can do to prolong the life of your tool is to learn the limits of the tool set. No matter how well you generally maintain it, using it abusively once will break it.

    You’ve got yourself a fine little knife, I hope it serves you well for years to come.















  • Let me expand a little bit.

    Ultimately the models come down to predicting the next token in a sequence. Tokens for a language model can be words, characters, or more frequently, character combinations. For example, the word “Lemmy” would be “lem” + “my”.

    So let’s give our model the prompt “my favorite website is”

    It will then predict the most likely token and add it into the input to build together a cohesive answer. This is where the T in GPT comes in, it will output a vector of probabilities.

    “My favorite website is”

    "My favorite website is "

    “My favorite website is lem”

    “My favorite website is lemmy”

    “My favorite website is lemmy.”

    “My favorite website is lemmy.org

    Woah what happened there? That’s not (currently) a real website. Finding out exactly why the last token was org, which resulted in hallucinating a fictitious website is basically impossible. The model might not have been trained long enough, the model might have been trained too long, there might be insufficient data in the particular token space, there might be polluted training data, etc. These models are massive and so determine why it’s incorrect in this case is tough.

    But fundamentally, it made up the first half too, we just like the output. Tomorrow some one might register lemmy.org, and now it’s not a hallucination anymore.