your eyes do not deceive you; this is a new issue on an actual Friday! I’m sick at home, so let’s cheer up with some machine learning!
The Latest Fashion
- This website explains any part of any paper that you highlight and allows follow-up questions.
- The Follower connects open camera data with AI to find live footage of influencers in the wild. Yes, it’s as creepy as it feels.
- I’m low-key obsessed with Stable Diffusion; here is a collection of open prompts.
Got this from a friend? Subscribe here!
My Current Obsession
This week I was off to Athens for me to attend the AI for Natural Disaster Management focus group of the ITU. It’s a UN thing trying to figure out how we can best apply machine learning methods to predict and mitigate natural disasters and the impacts of climate change. Really cool to be involved, and this time, they held a workshop accompanying it. I moderated session 2 on AI for monitoring and detection, and it went really well!
I also got a bunch of new Python calls for participation for pythondeadlin.es, namely:
- PyCon Bolivia
- Pydata Eindhoven
- PyCascades in Vancouver
- PyCon US
And finally, I contracted a cold on all this travelling. My rapid test is negative, I’m boosted, so I hope it’s alright and it’s not a false-negative. But so far, it feels like just a lot of stress and a cold conspiring against me.
Thing I Like
I’m still flying too much, so another shoutout to my noise-cancelling headphones. I’m surprised I ever flew without considering how loud planes are.
Hot off the Press
I have some great ideas for new videos, but now I don’t have a voice…
So please enjoy a classic: Never Include these Data Science Projects on a Resume
Machine Learning Insights
Last week I asked, “What is the concept of inductive bias?”, and here’s the gist of it:
Inductive bias is a term used to describe people’s tendency to prefer one thing over another in the absence of any information indicating a difference between the two things.
In machine learning, inductive bias is the tendency of a learning algorithm to search for certain patterns or rules. Algorithms with an inductive bias search for rules that have been found before.
The reason we can make different kinds of generalizations about the world around us is that induction is a useful heuristic that allows us to reach conclusions about the world without having to examine each and every instance of that object or event. So then, the inductive bias allows us to make generalizations about the world based on only a few observations of each object or event. There are many different kinds of inductive biases, but basically, they are a way for us to include expert knowledge into a learning algorithm, like physics-informed neural networks.
Data Stories
One of the most mind-blowing facts I started thinking about after reading Ender’s Game was realizing there is no “up” in space.
So how do we actually decide which way a telescope image should be rotated and published?
For many telescopes, the next question revolves around the colours. Telescopes often record images outside of the visible spectrum, but it’s where the most interesting pieces of data live. This article by The Verge has some really lovely “flip-book” images comparing palettes from different laboratories, and it’s beautiful!
Image Credit: NASA, ESA, and The Hubble Heritage Team (STScI/AURA) / Acknowledgment: N. Smith (University of California, Berkeley)
Question of the Week
- What is the curse of dimensionality, and how do you address it?
Post them on Twitter and Tag me. I’d love to see what you come up with. Then I can include them in the next issue!
Tidbits from the Web
- Keyboard shortcuts are great for productivity; here’s a list of apps on every operating system!
- The Taipei 101 tower was built to be “earthquake proof”, and now we have video of the counter-weight working hard to counteract the movement of the ground.
- I didn’t know prompt insertion attacks are a thing, but this is how you hack Dall-E 2.