Join 📚 Josh Beckman's Highlights
A batch of the best highlights from what Josh's read, .
More capable models can better recognize the specific circumstances under which they are trained. Because of this, they are more likely to learn to act as expected in precisely those circumstances while behaving competently but unexpectedly in others. This can surface in the form of problems that Perez et al. (2022) call sycophancy, where a model answers subjective questions in a way that flatters their user’s stated beliefs, and sandbagging, where models are more likely to endorse common misconceptions when their user appears to be less educated.
We Need to Tell People ChatGPT Will Lie to Them, Not Debate Linguistics
Simon Willison
The goal is not to fit in. If anything, it’s to amplify the differences, what doesn’t fit, the special characteristics unique to how you see the world. Instead of sounding like others, value your own voice. Develop it. Cherish it. The reason to make art is to innovate and self-express, show something new, share what’s inside, and communicate your singular perspective.
The Creative Act
Rick Rubin
Tree-sitter’s parsers already let us efficiently parse the code that our users upload. For instance, the Tree-sitter parser for Python produces a concrete syntax tree (CST)
Introducing Stack Graphs | the GitHub Blog
github.blog
...catch up on these, and many more highlights
Readwise