Finnish Center for AIMar 13, 2024

In the latest FCAI newsletter:
📢 Apply now for the new Finnish PhD program in AI
📓 Collaborative AI and Modeling of Humans workshop at #aaai2024
💊 Machine learning for personalized medicine
🤖 CfP for AI Day + #NordicAIMeet

Read here: https://mailchi.mp/1588702a6f71/fcai-newsletter-17370901

patalt Mar 4, 2024

#AAAI2024 at dawn, boarding at dusk 😎

Had an amazing time in Canada full of interesting workshops and talks, great interactions and fresh pow!

1st photo: with @avandeursen in front of our poster (w/ Mojtaba Farmanbar and @cynthiacsliem)

2nd photo: on the slopes of Cypress (ski resort near Vancouver)

Show threadUlrich JunkerFeb 26, 2024
Nevertheless, I felt that there have been less presentations at #AAAI2024 about classic #AI topics such as #ConstraintSatisfaction, #KnowledgeRepresentation and #Reasoning, #Planning, #Scheduling, #Search and so on. I only hope that these topics will not end like rule-based systems, which are nowadays used in many business applications, but largely ignored in academia. On the positive side, there has been a good number of presentations about economic paradigms and game theory.
Ulrich JunkerFeb 26, 2024
Some impressions from my remote participation at #AAAI2024: #LLMs have become inescapable, but many researchers are just using them as a building block. Others do some praise of LLMs, then tell what they are really interested in, and then continue their praise of LLMs. In particular, LLM critics is sometimes wrapped into praise. Nevertheless, the panel about impact of LLMs has been very good and the talk by #PascaleFung about LLM hallucinations clearly exposed the limitations of those systems. ⤵️
patalt Feb 13, 2024

My newest blog post, ECCCos from the Black Box, provides an accessible and high-level overview of our upcoming #AAAI2024 paper.

https://www.paltmeyer.com/blog/posts/eccco/

Free of any technical details, maths or code (although you will find links to our Julia packages )

Patrick Altmeyer - ECCCos from the Black Box

ECCCo is a new way to generate faithful model explanations through counterfactuals that are as plausible as the underlying model permits. Companion post to our AAAI 2024 paper.

Patrick Altmeyer
patalt Jan 17, 2024
Anyone @UBCCS who might be interested in Julia  and Trustworthy ML? I will be in Vancouver in February for #AAAI2024 and would be happy to chat about our [Taija](https://github.com/JuliaTrustworthyAI).
Taija

Home for repositories of the Taija (Trustworthy Artifical Intelligence in Julia) project. - Taija

GitHub
Show threadpatalt Jan 9, 2024

We ran large experiments including various models and datasets from different domains and find that the performance of ECCCo is robust.

All of our code is open-sourced and built on top of our CounterfactualExplanations.jl  library (will eventually be absorbed): https://github.com/pat-alt/ECCCo.jl.

In February, I'll present this research at #AAAI2024 in Vancouver. Let me know if you're planning to attend and would like to connect.

7/n

GitHub - pat-alt/ECCCo.jl

Contribute to pat-alt/ECCCo.jl development by creating an account on GitHub.

GitHub