Goodhart's Law and Overfitting

Author

Bobby Brazen
November 11, 2022

Here's your daily briefing:

Also we still have a fever, so forgive us for the lighter, more link-heavy post!

  • Research scientist Jascha Sohl-Dickstein shared his first blog post, which is about Goodhart's law ("when a measure becomes a target, it ceases to be a good measure") as applied to machine learning and the problem of overfitting.

Too much efficiency makes everything worse: overfitting and the strong version of Goodhart’s law

This blog is intended to be a place to share ideas and results that are too weird, incomplete, or off-topic to turn into an academic paper, but that I think may be important. Let me know what you think! Contact links to the left.

https://sohl-dickstein.github.io/2022/11/06/strong-Goodhart.html

  • Check out this Airbnb we just rented:

Just kidding, that's a sample from the Lexica Aperture model 😮 :

  • We agree, Chris. If you're thinking about learning some valuable data science skills, definitely do not go into debt before you check out these FREE resources:

  • The news has been proven to be bad for your mental health. This 12 year old may have done more for collective well-being than the next 100 adult "mental health advocates":

  • Could this new prompt-based search engine by Metaphor be part of a breakthrough in internet search? Google has certainly held the mantle for long enough...

  • Wild to think of this a year or two from now:

  • We're quite curious to know what Sam is detecting on "radar through the fog" that the rest of us haven't even thought about 🤔

  • And now, some generative art:

We'll see you tomorrow!