Categories

Develop a prototype app that assesses and promotes fairness for data workers (in participative projects, volunteering initiatives, online gigs), leveraging the FINDHR toolkit to ensure non-discriminatory practices.

Use the new Swiss models inside of an open source game engine. Inspire players and indie developers to be more aware of the AI powering their experience - especially those who are interested in Switzerland's culture, languages, and history.

Accelerate chatbot interactions with real-time interpretation of facial expressions.

Apply AI techniques to enhance transit detection in space telescope data

A challenge inspired by the "Bern in Minecraft" project at Data Hackdays BE. Our task is to use AI to evaluate and improve the LIDAR 3D scans of environments in the City of Bern and other locations in the Canton. Such digital twins could be used for art and video games, as well as more 'serious' uses like city planning and emergency services. Demonstrate that AI can learn from and augment our capabilities in making use of this complex data source.

Build your own "baby" Large Language Model with help from "momma" LLM

LLM's are changing how we browse the Internet. Create the future of web browsing using LLM and open tools like: https://staan.ai/ or https://openwebsearch.eu/ .

This challenge to tap into public data sources is completely AI-generated.

Motivated by the ideas in this challenge, we invited a local musician to perform a set during the Hackathon. This was the first live act of CyberGwen, who has already a growing audience online. The visuals in the background from open source shader artists complemented the act. It was epic. Find the recording here:

Use a hardware robotics platform in combination with LLMs to replicate fluid and "natural" motion in the environment, or some set of interesting interactions that can be modelled with prompts. You could also propose a more versatile learning tool, along the lines of Teachable Machines. An open source robot will be available at the hackathon for you to test your solution on, and there are several companies interested in the use cases you explore.

It is easy to spot an AI generated text. Just look for the Eszett: ß. Germans love it and Swiss won't have it. This is just one example of how Swiss localisation in LLMs not given. There are word frequencies that favor the high German language. This is because most LLMs are fed with a lot German texts. Is this actually an issue? How can we untrain, retraint, train LLMs to actually "speak" Swiss german.

Nachhaltige KI-Infrastruktur für verantwortungsvolle Schweizer digitale Dienstleistungen