We want to do both. In finance, highly regulated industry, understanding how models work is critical. In addition, mech interp will allow us to understand which current or new architectures could work better for financial applications.
Thank you for reading. One of the main reasons we've written the paper is to help with model validation of LLM usage in our highly regulated industry. We are also engaging with regulators.
The industry at the moment is mostly using closed sourced vendor models that are very hard to validate or interpret. We are pushing to move onto models, with open source weights and where we can apply our interpretability methods.
Current validation approaches are still very behavioral in nature and we want move it into mechanistic interpretation world.
Thank you. Agreed, we are exploring different ways to apply these interpretability methods to a wide range of transformer based methods, not just decoder based generative applications.
Paper introduces AI explainability methods, mechanistic interpretation, and novel Finance-specific use cases. Using Sparse Autoencoders, we zoom into LLM internals and highlight Finance-related features. We provide examples of using interpretability methods to enhance sentiment scoring, detect model bias, and improve trading applications.
Our paper introduces AI explainability methods, mechanistic interpretation, and novel Finance-specific use cases. Using Sparse Autoencoders, we zoom into LLM internals and highlight Finance-related features. We provide examples of using interpretability methods to enhance sentiment scoring, detect model bias, and improve trading applications.
My younger daughter, who is pretty good at making games in Scratch is not that interested in jumping into text/code based programming. I do think Scratch makes things a lot easier and text based programming is not thar appealing to kids. I will try to start her with Pygame but even that might make it seem very arcane and not very visual.
PICO-8 [1] might be a good choice. I always consider that kids want their friends to try their games, and so being able to easily distribute to the web is awesome. They can link their game from the site [2], or with some parental help they could even serve them from their own website, which could be very empowering for a kid.
Another vote for Pico-8. It’s such an incredible little package.
The simple IDE and forced constraints makes you really focus on the basics like fun gameplay loops and minimal graphics, and the fact you can do the code, graphics, sound effects, and music all in one little program is a really smart way to teach you all different aspects of game dev.
It’s also so ridiculously easy to share your creations, outputting a simple HTML/JS combo that works perfectly on mobile and desktop even if they don’t own the program.
My daughter (12) got her start in Scratch, but had a hard time jumping into Python. She's enjoyed GBStudio, and has made a number of small games in that environment, and enjoys loading them onto a flash cart to run them on an old physical Gameboy.
https://code.org/student/middle-high has an environment that's scratch-like i the UI but the code is blockified JavaScript. (If you try to transpile a Scratch project to JavaScript using normal tools, you get an unmanageable JSON monstrosity)
It hides some of the text syntax, while still being an onramp to text-based programming.
Terrence Tao has a very healthy view of what AI can and cannot achieve shorter term. It is refreshing to hear more grounded views from a top mathematician who is clearly well versed in the topic.
