Introduction

Pycon 2024 took place May 15-23 in Pittsburgh, PA. I went to the Saturday and Sunday days May 18-19. There were many tracks and open spaces happening simultaneously. Open spaces were round tables like birds of a feather where people discussed arbitrary topics. There was even a track in Spanish. I went to mostly lectures on Saturday. Then, I went to mostly open sessions on Sunday.

Memorable Lectures

Sim Willison Keynote

This talk was mostly about large language models (LLMs). He went a little into copyright issues in the training data. Specifically, he mentioned that Meta’s LLaMa model had some copyright issues. 4.5% of LLaMa’s training dataset Sampling was books according to LLaMa: Open and Efficient Foundation Language Models. An author, Sarah Silverman, has sued Meta and OpenAI for that. It wasn’t mentioned in the talk, but there was a similar case with the New York Times sueing OpenAI and Microsoft. OpenAI’s CTO, Mira Murati, mentioned that they use “publicly available videos”. But it is against YouTube’s policy to train on their videos.

Another thing that he mentioned is that LLMs don’t have a great way to objectively evaluate them. He says that the best way is to qualitatively compare them. He jokingly calls it vibes. There’s a tool called lmsys that does a side-by-side comparison of two different models with the same prompt.

Jacob Lapenna: Open Source Industrial Control: Turning 2,800 Tons of Metal with Python and Flask

This talk covered some work managing a large piece of power generating infrastructure. The speaker is an electrical engineer. His project improved the workflow of maintaining what is a giant DC motor. Overall, it was very interesting to see someone putting together mechanical, electrical, and software skills.

Notes from Open Sessions

Dev Ex

There was talk about shadow ops - developers building their own infrastructure with little involvement of IT. Someone mentioned using localstack to improve AWS testing.

Ricing

Some folks showed off their desktop environment and their tools. There were a lot of vim enthusiasts. One member mentioned using a tool called btop. Someone showed off a song with bespoke synth.

AI Tools

1. cursor: an IDEA where LLM is at the main feature.
2. HASH: an AI project that Joel Spolsky supports
3. notion: AI notebook arc: an AI bot that tracks investments
4. quivr: a chatbot trained on custom user data
5. ShellGPT: A command line tool with LLM integration
6. askmarvin: a toolkit for building natural language interfaces
7. descript: a tool that allows editing podcast audio via text to produce a new audio track.
8. VOCALOID6: AI generated singing.
9. BandLab: music producing software

Rust

1. Online Intro Rust Book
2. rustlings: small rust exercises
3. Learn Rust With Entirely Too Many Linked Lists
4. Rust by Example
5. Evcxr: a rust interpreter that may be used in Jupyter

Talks to watch

• Anthony Shaw: Unlocking the Parallel Universe: Subinterpreters and Free-Threading in Python 3.13. 19 May, 02:30 PM - 03:00 PM Track 1 (Ballroom A)
• Jay Miller: Keynote. 7 May, 09:45 AM - 10:30 AM Main Stage
• Reuven M. Lerner: Times and dates in Pandas. Fri 17 May, 2024 - 12:30 PM-01:15 PM Track 2 (Ballroom BC)
• Jules Poon and Ken Jin Ooi: How two undergrads from the other side of the planet are speeding up your future code. 19 May, 01:45 PM - 02:15 PM Track 1 (Ballroom A)
• Kate Chapman: Keynote. 19 May, 09:00 AM - 09:40 AM Main Stage
• Kevin Kho and Kevin Kho: Speed is Not All You Need for Data Processing. 18 May, 01:45 PM - 02:15 PM Track 4 (Hall C)
• Neeraj Pandey and Manoj Pandey: Visual Data Storytelling with Blender and Python. 18 May, 10:30 AM - 11:00 AM Track 1 (Ballroom A)
• Sebastián Flores: Python and Data Storytelling to the Rescue: Let’s Avoid More Deaths by PowerPoint. 18 May, 12:00 PM - 12:30 PM Track 1 (Ballroom A)
• Sumana Harihareswara: Keynote. 19 May, 03:15 PM - 03:55 PM Main Stage

Source for Post

Comments