I am trying to introduce my students in software engineering to LLMs. I want them to have a good idea of what it is like to work with them, or to build tools on top of them.  What is the best option currently?

It seems that I have almost entirely replaced my use of Grammarly and LanguageTool with ChatGPT (GPT4). The ability to instruct the AI on what to do in different parts, --ignore latex comments, keep the structure, make abstract concise, make introduction simple etc.--are a real game changer for me. However, I find its ability to write code I want *extremely* lacking!.

Valiant's general context free parser is one of the few very well known general context free parsers. Its claim to fame is that given any context-free grammar, it can parse any strings in that grammar in less than O(N^3) time, by leveraging advances matrix multiplication. Unfortunately, lucid explanations of what Valiant's algorithm does is somewhat hard to find. I have just written a live notebook post that explains the Valiant's parser, and lets you play with the algorithm directly.

https://rahul.gopinath.org/post/2023/03/16/valiant-parser/

#parsing #context-free

An unforeseen (or perhaps foreseeable) issue; Most foot guns I have learned to avoid in Python are present in GPT4 generated code, and it is not immediately obvious anymore. Chased a surprising bug and found 
             C = [[{}] * m for _ in range(m)]
followed by 
         C[i][j][k] = "V"
buried deep. Because I trusted GPT4 (it aught to be good at these things) I failed to recognize this pattern :/

It is very mesmerizing to get chatGPT+ (GPT4) to write code for you. It keeps making different mistakes every time, and inventing newer explanations!

To those who are struggling with Microsoft's outlook, please give #mutt http://www.mutt.org/ a try. It works very well with outlook (beyond some configuration), has sane searching capabilities, and does exactly what you want it to do. I have been using it to survive email in #usyd.

A few important tips for folks running #mastodon instances (also for people who run their own microbloging instances like #ktistec) https://denise.dreamwidth.org/91757.html

Has any one gotten #chatGPT3 to do a literature survey correctly? It summarizes the research in many fields reasonably well, but it always makes up papers. I couldn't get it to cite even a single real paper.