📢Open-source #AI has a lot of pros ... but also cons. as discussed in our recent preprint 👉Balancing Transparency and Risk: The Security and Privacy Risks of Open-Source Machine Learning Models 👉arxiv.org/abs/2308.09490 Still a lot work to be done, technologyreview.com/2023/08/17/107…

LLM's might have capabilities but causality isn't one of them!! This is one of my favorite papers and I am happy it is now published in Transactions on Machine Learning Research. We show how LLM's are "causal parrots" although they may seem causal at face value. With Matej Zečević Moritz Willig Kristian Kersting

This will be an exciting AAAI 2024 workshop: LLMs and causality? Are they causal 🦜? What does that imply for reasoning and planning? Looking forward

We convert popular knowledge graph embedding (KGE) models such as ComplEx into generative models of triples that 1️⃣ better scale training to very large knowledge graphs 2️⃣ make reliable predictions with logical constraints 3️⃣ support sampling in our #NeurIPS2023 oral!

😂🔝Manuel Brack

Thanks to everyone who joined our oral 1C session or discussed later at the poster for Characteristic Circuits at #NeurIPS2023 Check out our paper at: arxiv.org/abs/2312.07790 Kristian Kersting

Happy and proud to have Massachusetts Institute of Technology (MIT)'s Tomaso Poggio TOMASO POGGIO give a talk at Centre for Cognitive Science and hessian.AI ! TU Darmstadt

We’ve also shown at UAI23 w/ Fabrizio Ventola Steven Braun Kristian Kersting that overconfidence isn’t unique to NNs, but also practically occurs in probabilistic circuits - tractable models. We’ve shown how we can combat this through tractable uncertainty in PCs: proceedings.mlr.press/v216/ventola23…

Please consider submitting your works on #Tractable #Probabilistic #Models at the #TPM2024 workshop @ #UAI2024 UAI 2025 --- still ~10 days left 👉 …able-probabilistic-modeling.github.io/tpm2024/cfp/ 🗓️Deadline: May 27th, 2024 AoE #probabilistic #circuits #normalizing #flows #pgm #trustworthy #AI

We learn more expressive mixture models that can subtract probability density by squaring them 🚨We show squaring can reduce expressiveness To tackle this we build sum of squares circuits🆘 🚀We explain why complex parameters help, and show an expressiveness hierarchy around🆘

🙏for the kind words. I am standing on the shoulders of giants 🦾 xLSTMs are amazing and work really well for multivariate time series. All the credit to the amazing team Maurice Kraus @ NeurIPS Felix Divo Devendra Singh Dhami | देवेंद्र सिंह धामी 🇮🇳

Our recent workshop paper puts Vision-Language Models like #GPT4o to the test with Bongard problems🧩 We show that VLMs struggle to identify concepts that are quite intuitive to humans. Still a long way to go for human-like visual reasoning! 🤖🧠 arXiv: arxiv.org/abs/2410.19546

"You can't chatbot your way out of climate change". Great points from Sasha Luccioni, PhD 🦋🌎✨🤗 from Hugging Face on BBC!

Aktuelle #KI-Modelle bestehen nicht KI-Benchmarks aus den 1960iger Jahren 😤 Tolle Zusammenarbeit mit Antonia Wüst Tim lukas helff Devendra Singh Dhami | देवेंद्र सिंह धामी 🇮🇳 ConstantinRothkopf heise.de/news/Vision-La…

Less than two weeks to submit your papers on: 📈 #lowrank adapters and #factorizations 🧊 #tensor networks 🔌 probabilistic #circuits 🎓 #theory of factorizations to the first workshop on connecting them in #AI #ML at AAAI please share! 🔁 👇👇👇 april-tools.github.io/colorai/

Folks, how would you go about this?

the #TPM ⚡Tractable Probabilistic Modeling ⚡Workshop is back at UAI 2025 #UAI2025! Submit your works on: - fast and #reliable inference - #circuits and #tensor #networks - normalizing #flows - scaling #NeSy #AI 🕓 deadline: 23/05/25 👉 …able-probabilistic-modeling.github.io/tpm2025/

Happy that our paper on how to make HPO & NAS (& BO) more interactive via Probabilistic Circuits has been well-received! We leverage PCs' flexible inference to incorporate human feedback on arbitrary subsets of hyperparameters anytime during optimization! Check it out 👇