Yusuke Oda

Turn Python into LaTeX equations and pseudocode! How cool is that? latexify.predicate.jp via @odashi_en

I deployed LaTeXify Web: converts Python functions into LaTeX algorithmic!

Released latexify v0.3.1. This version involves lots of updates. Some interesting features:

Tuna. twitter.com/3DVconf/status…

Code/comment generation was ones of my research areas 8-10 years ago, but they were very difficult and I finished only a limited work. I wonder if I were a CS student this year😉

👉 github.com/odashi/davinci…

davinci-functions implements several utilities to ask GPT to do something in Python:

Implemented davinci_functions.function(prompt) that synthesizes Python callable from the given prompt. github.com/odashi/davinci…

The opposite direction: given a function, let GPT describe its behavior.

Interesting. I wrote a function to synthesize a function using the result of GPT, and demonstrates how we could inject malicious prompts. My prompt involves a bug around os.environ (it is definitely not intended) but the result works successfully because GPT corrected it.

Interviewer: "Please write the solution of the two-sum." Me: "Yeah, GPT, do it." GPT: "Yeah."

The picture above doesn't show several rows for f(27) (because I accidentally scrolled the tab). This is the output at the beginning:

Syhthesizing a Python function by GPT and run it on Python!

davinci-functions, a library to obtain *Python literals* from GPT, is now available on PyPI: github.com/odashi/davinci…

This works as follows:

Just a fun thing: I wrote a very simple wrapper that generates Python list by asking GPT-3: github.com/odashi/davinci…

With recent AI progress, it’s never been easier to build AI-driven apps! But does your app/system work as expected? Or does it serve up false facts, incoherent gibberish, or toxic content? Introducing *Critique*, a quality-control tool for generative AI! docs.inspiredco.ai/critique/index…

@vbkaisetsu @kampersanda Implementation available in Rust! github.com/daac-tools/daa…

My work has been published to "Software: Practice and Experience", which describes various ways to construct *very* efficient pattern matching machines. Joint work with @vbkaisetsu and @kampersanda. Read-only ver: onlinelibrary.wiley.com/share/author/B… Draft (arXiv): arxiv.org/abs/2207.13870

Hello World! We are Inspired Cognition, and our mission is to make it easy for everyone to efficiently and effectively build *reliable* AI-driven systems. We have a few announcements coming up soon, so watch this space 👀