Focus AI #105
#Hacking #Mastodon #Python #Go
Focus AI, the latest AI news in 3 minutes
Edition #105 - November 25th 2023
Feel free to subscribe or to share it.
This week, only 3 news got my attention:
🧨 This 'thermal attack' can read your password from the heat your fingertips leave behind
♟️ Champion Model Is No Go
🌳 When Trees Outdo Neural Networks
And about code:
[Repo 📊] Nanny v0.8.0
[Podcast 🎙️] Mastodon for Python Devs
[Code 🐍] What’s New In Python 3.12
== News ==
🧨 This 'thermal attack' can read your password from the heat your fingertips leave behind
Computer security researchers say they've developed an AI-driven system that can guess computer and smartphone passwords in seconds by examining the heat signatures that fingertips leave on keyboards and screens when entering data.
Called ThermoSecure, researchers at the University of Glasgow's School of Computing Science developed the system to show how the falling price of thermal-imaging cameras and increasing access to machine-learning and artificial intelligence (AI) algorithms are creating new opportunities for what they describe as thermal attacks.
♟️ Champion Model Is No Go
A new algorithm defeated a championship-winning Go model using moves that even a middling human player could counter.
Researchers at MIT, UC Berkeley, and the Fund for Alignment Research trained a model to defeat KataGo, an open source Go-playing system that has beaten top human players.
In the usual approach, the model plays itself and learns from all moves. In the authors’ version, the model played against a fixed KataGo model and learned only from its own moves, learning to exploit holes in KataGo’s strategy rather than becoming a conventionally savvy player.
🌳 When Trees Outdo Neural Networks
While neural networks perform well on image, text, and audio datasets, they fall behind decision trees and their variations for tabular datasets. New research looked into why.
What’s new: Léo Grinsztajn, Edouard Oyallon, and Gaël Varoquaux at Inria Saclay Centre and Sorbonne University trained a variety of neural networks and tree models on tabular datasets. Performance on their tabular data learning benchmark revealed dataset characteristics that favor each class of models.
Key insight: Previous work found that no single neural network architecture performed best on a variety of tabular datasets, but a tree-based approach performed better than any neural network on most of them. Training and testing different models on many permutations of the data can reveal principles to guide the choice of architecture for any given dataset.
== Code & Tools ==
Let’s dive in into more technical news.
[Repo 📊] Nanny v0.8.0
In the past 6 Months hey have added:
- Regression Support (with Direct Loss Estimation)
- Command Line Interface
- Database exports and cloud integration
- Official docker container
- A ton of new univariate data drift tests (Jensen-Shannon Distance and L-infinity distance)
- A bunch more improvements
[Podcast 🎙️] Mastodon for Python Devs
Wondering what Mastodon is all about? More importantly, what does it offer Python developers and other open source folks compared to Twitter? There is a huge amount of interest in the tech community about what's happening at Twitter and whether they should expand to or even move to a new location.
On this episode you'll meet Gina Häußge, Simon Willison, and Juan Luis Cano Rodríguez. We're going to have a great time talking about the technology and the culture of Mastodon from this Python perspective.
[Code 🐍] What’s New In Python 3.12
This article explains the new features in Python 3.12, compared to 3.11.
#Hacking #Mastodon #Python #Go
A big thanks to our sources: https://jack-clark.net/, https://www.actuia.com/, https://thevariable.com/news/, https://techcrunch.com/, https://read.deeplearning.ai/the-batch/
What about you? Have you noticed something else?
Don’t miss the next news!
Have a good week-end,
Maxime 🙃 from Toulouse, France with 🌺