Kai boosted

Stable Diffusion 2.0 introduces depth2img, which is like img2img but uses depth map estimates to help guide the generation. This should help in situations where img2img doesn’t understand how to segment foreground and background, creating weird artifacts.

Show thread
Kai boosted
Kai boosted

🧵1/ The Netherlands is the place to be if you want to study the #government's misuse of #AI in a democratic country.

A week or so ago there was some reporting on the Top 400, a #predictive #crime initiative by the City of #Amsterdam. They algorithmically create a list of the 400 youngsters they believe are most likely to commit serious crimes. The #police actively surveills them (including house visits!), even if they haven't committed any crimes at all.


Kai boosted

RT @biorxiv_bioinfo@twitter.com

OpenFold: Retraining AlphaFold2 yields new insights into its learning mechanisms and capacity for generalization biorxiv.org/cgi/content/short/ #biorxiv_bioinfo

🐦🔗: twitter.com/biorxiv_bioinfo/st

Kai boosted

Here's a 1-toot, 2-step guide to Mastadon for those coming from Twitter:

1⃣ The choice of server isn't important. (You can follow people on other servers, and if needed you can change server later.) Don't get analysis paralysis! I recommend any of fosstodon.org, qoto.org, sigmoid.social, mathstodon.xyz depending on how you want to brand yourself.

2⃣ Everyone is leaving #introduction posts to get started. You should too!

(...yes, they're called Toots here. 😄🦣)

Kai boosted

Time to kick off this Mastodon thing with a #caturday post!

And now that I have your attention! ;) I write open-source #JAX and #PyTorch software for neural networks, different equation solvers, numerical methods, scientific computing, ... : github.com/patrick-kidger

Currently employed at #Google X, working on CompBio problems. Previously PhD @ University of #Oxford , studing neural ODEs!

( #introduction :) )

Kai boosted

Introduction time for me, an funded science project called .

TL;DR: we build a (bioimage.io) to facilitate based image data processing on data obtained in a context. Why? Because it is important to make the most out of all the amazing data out there!!!

Our 10-partner consortium will build an open, accessible, community-driven repository of pretrained models and develop services to deliver these models to life scientists, including those without substantial computational expertise. will provide direct support and ample training activities to prepare life scientists for responsible use of methods. Additionally, will drive community contributions of new models and interoperability between analysis tools. will also facilitate and public aimed at providing state-of-the-art solutions to unsolved image analysis problems in life science research.

brings together /#ML researchers, developers of image analysis tools, providers of European-scale storage and compute services, and European life science -- all united behind the common goal to enable life scientists to benefit from the untapped, tremendous power of AI-based analysis methods.

Check out our model zoo at bioimage.io…

Kai boosted

Your reminder that a large, careful study showed that the number one factor that separates high-performing software teams from the rest is psychological safety: everyone feeling comfortable asking questions and providing constructive criticism. forbes.com/sites/cyrusfarivar/

Kai boosted

It seems that the etiquette for this site is for your first post to be a self introduction. Oops. Anyway, for anyone who wants to know if they should follow me, I'll say: I don't post much, but when I do, it's usually about Bayesian machine learning, or my ML textbooks (probml.github.io/pml-book/), or my JAX software (stay tuned for a major update on that front). So if that's your jam, let's chat!

Kai boosted

To add to the above:

Since there is no #algorithm it is really important that you #boost interesting posts by others.

A star (liking) does nothing to help increase a post's visibility beyond the original author's followers.

I know it is strange coming from 🐦, but it really is the only way to increase a post's audience.

So #BoostAllTheThings (if you like them) and you will notice others will do the same to your content. Here we're all in it together!

#BeTheAlgorithm #BoostsWelcome #FediTips

Show thread
Kai boosted

I am delighted to announce the release of “dynamax”, an open source library for dynamic state space models in JAX. It supports inference and learning in HMMs, Linear Gaussian SSMs, as well as non-linear and non-Gaussian SSMs. See github.com/probml/dynamax for the code.

Kai boosted

📚 What are you top general Machine Learning books ?

I would say:
📘 Deep Learning - Goodfellow, Bengio and Courville
📗 Artificial Intelligence: A Modern Approach - Russel & Norvig
📙 Pattern Recognition and Machine Learning - Bishop

#IA #ML #MachineLearning #book #books #recommendations #academia #question #discussion #media

@solalnathan Clearly also the book series in Probabilistic Machine Learning by @sirbayes - see probml.github.io/pml-book/

Kai boosted


Understanding ML (cs.huji.ac.il/~shais/Understan) by Shai Shalev-Shwartz and Shai Ben-David.

Bayesian Data Analysis (stat.columbia.edu/~gelman/book) by Andrew Gelman, John Carlin, Hal Stern, David Dunson, @avehtari and Donald Rubin.

Mathematics for Machine Learning (mml-book.github.io) by @deisenroth, Aldo Faisal and @chengsoonong.

I haven't read it yet, but it looks great:

Probabilistic Numerics (probabilistic-numerics.org/tex) by @PhilippHennig, @maosbot and Hans Kersting.

Kai boosted
Kai boosted

"participants who had access to the AI Assistant were more likely to introduce Security Vulnerabilities for the majority of Programming Tasks"


Kai boosted

Dear Twitter ex-pats. When we all flounced out of Twitter loudly proclaiming we were going to Mastodon this sent a message to trolls, nazis that they could set up servers & come here.

If anyone gets harrassed with hatespeech pls note that that is not the culture of Mastodon. Please report & block those users. If you notice a pattern from the same weird domain, block that domain, write a post of the server name w/ a hashtag fediblock so your server admin can follow up & block it for everyone.

Kai boosted
Kai boosted

NeRFs (Neural Radiance Fields) are neural networks that can generate 3D scenes from 2D images.

NVIDIA's Developer Blog has a great write-up explaining how they work.


#AI #MachineLearning #AIArt

Kai boosted
Show more
Qoto Mastodon

QOTO: Question Others to Teach Ourselves
An inclusive, Academic Freedom, instance
All cultures welcome.
Hate speech and harassment strictly forbidden.