🥔 Potatoes sprout a smile on my face
Late to the Party 🎉 is about insights into real-world AI without the hype.
Hello internet,
I’m on a train halfway through Germany. It truly has been an eventful week! Let’s talk about some machine learning!
The Latest Fashion
- Keras is getting a Pytorch and Jax backend!
- The internet is going wild over Gzip beating transformers in text classification!
- Wikipedia uses a bunch of Xgboost models to detect vandalism!
Got this from a friend? Subscribe here!
My Current Obsession
Ok, big news!
You all knew I was working on a Skillshare course about chatGPT, and it’s live now!
I teach content creators and creatives how to generate ideas, build on their existing content library, the dangers of chatGPT, and even co-creation and the future of AI and content creation. Very proud of this course, and I hope it cuts through the noise.
I also published an ebook about “ChatGPT for Creatives and Content Creators”, which is 35 pages of deep dives into different ways to up your prompt engineering game.
Then I took off to Berlin to attend the International Union of Geodesy and Geophysics, a conference for geoscientists. I attended a lovely session on AI in natural disaster management. I then had some great meetings with the people at Fraunhofer-HHI to talk more about the developments of digital technologies in geoscience. I posted about it here!
Meta also restricted my access to Threads, which sucks. But I had fun while it lasted.
An Update on Sponsors
People have approached me about sponsoring this newsletter.
This newsletter has reached over 2,000 inboxes over 2 years, and I have never taken sponsorships. This newsletter will stay independent and really just bring you projects and news I find interesting.
You can, of course, upgrade to support this publication. I appreciate those of you who support this work directly!
But in the end, I'm just happy to send these issues to people who care about the actual workings of machine learning and AI beyond the shiny chatGPT FOMO (fear of missing out).
Thanks for being here!
Thing I Like
The weekend in Berlin was scalding hot. Luckily I was hanging out with my friend from Spain, and we got some classic Spanish fans to keep us cool without batteries. Really makes you feel better in this heat.
Hot off the Press
I wrote a blog post about optimising effective ChatGPT Prompts using the RTF format!
I also published a short video about the remote development plugin for VSCode.
My course on chatGPT for Content Creators is now live, and my book about chatGPT for Creative and Content Creators is also live!
I also co-wrote a blog post about community building called “Zero to not-Zero” with Meag Doherty.
In Case You Missed It
I shared FasterAI with my Linkedin audience, which you saw 15 months ago in this newsletter. My post on chatGPT failures, which you read five months ago, made rounds on Mastodon and Linkedin.
People have recently been reading my post about getting a machine learning job at ECMWF.
Machine Learning Insights
Last week I asked, Can you write an end-to-end machine learning model in 500 characters?, and here’s the gist of it:
import xgboost as xgb
from sklearn.datasets import load_wine
from sklearn.model_selection import train_test_split
X, y = load_wine(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y)
model = xgb.XGBClassifier().fit(X_train, y_train)
accuracy = model.score(X_test, y_test)
print(f"Accuracy: {accuracy}")
Data Stories
The Earth is a weird place.
The magnetism is wobbly and apparently sometimes flips. Our highest mountains and deepest oceans seem massive, but if Earth were shrunk down, it would be smoother than a billiard ball.
It’s fascinating.
The ground beneath our feet isn’t equally distributed either! Some places are heavier, and some are lighter, so they cause different gravitational pull.
That’s where the Potsdam Potato comes in!
It’s a visualisation of the pull the Earth has on you in different places on the surface!
It’s named the Potsdam potato because it was generated by the German research centre for Earth Science in a city called Potsdam. And it’s quite widely used as a global model of gravity!
Here’s the Potsdam Potato
Question of the Week
- Describe a good system to manage neural network configurations!
Post them on Mastodon 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
- Playing around with the world’s strongest magnet.
- This climbing battle was weirdly fun to watch!
- Will AI take your Job?
Jesper Dramsch is the creator of PythonDeadlin.es, ML.recipes, data-science-gui.de and the Latent Space Community.
I laid out my ethics including my stance on sponsorships, in case you're interested!