Pythia: Open-source framework for multimodal AI models | Hacker News Reader

Pythia: Open-source framework for multimodal AI models

code.fb.com

66 pointstilt18 comments

I had a project named Pythia as well in my previous company. It was a prediction service and I needed a name that was related to predicting future, hence I chose the name of Oracle of Delphi. I am guessing that's where even Facebook got the name from.

The name was so bad that no one got the reference. I changed the git repo to Sybill, alas that also never clicked.

Edit: was supposed to be a reply to `davrosthedalek`

Sybill Trelawny?

Yeah. Trelawney is more recognisable but doesn't have the same catch.

Oh, I thought it was the Cumaean Sybil. But I guess that is where Trelawney's name comes from.

Didn't know about her. Learned something new. :)

I made a prediction service called Pythia in 2009. It's not exactly an obscure name.

Not obscure but certainly not well known as well. Most are aware about the Oracle but not the name.

PS your name sounds Greek so it would definitely be more known in your circle I guess.

The project pages are notably lacking images or interesting examples. Here is the most recent paper showing what they use this system for: [1] Toward VQA Models That Can Read [VQA == Visual Question Answering]

[1] https://arxiv.org/pdf/1904.08920.pdf

It has example images with Q&A now:

https://github.com/facebookresearch/pythia

You can also try the colab notebook available at https://colab.research.google.com/drive/1Z9fsh10rFtgWe4uy8nv... to test the model in realtime.

Why do people not do a google search before they name things: http://home.thep.lu.se/~torbjorn/pythia81html/Welcome.html Edit: Also: https://en.wikipedia.org/wiki/PYTHIA

Yep, pretty sad to see Facebook guilty of this.

I named a project Pythia also. It's a pretty obvious choice for a python project. Name duplication is common. We will all live. No one will mistake this project for a particle accelerator sim.

You are assuming it wasn't intentional, which is not clear at this point (and probably never will be)

I recently learned about this piece of software when out for pizzas with a physicist friend of mine. I don't know if FB didn't know about it or didn't care.

Relatedly, people seem to love taking names from physics, too. Atom, electron, and hydrogen are the most frustrating ones. But there are a lot of libraries. It makes finding the actual physics information I'm interested in annoying.

When they themselves have repurposed the name, can we really be surprised that someone else chooses to do so? Perhaps there are outraged Greek scholars out there fuming at "those damn Physicists" for getting in the way of their search results.

It's like naming a project after an ancient God - you've sort of got to accept that there'll be some namespace clashes.

Looking forward to papers using pythia the AI library to analyze data simulated by pythia the physics library.

Reply on news.ycombinator.com