Energy use transparency 

A new personal energy use dashboard (available under Settings → Usage) shows your usage of AI models and the estimated environmental impact of AI models in Wh (watt-hours. This supports students and staff in making more evidence-based decisions on their use of AI and contributes to sustainability awareness within the university. Energy consumption is based on token costs and an input-to-output ratio of 9:1. The estimation is based on Samsi et al. (2023)External link and Luccioni et al. (2024)External link.  

Collaboration on personas and announcements 

Group owners can collaboratively create and edit AI personas, making it easier to design shared AI tools for courses and projects. Additionally, group owners can post announcements to all group-members. 

Expanded file formats and tagging ‘@’ 

Students and staff can upload .txt, and .md files directly into chats and personas, allowing to work with code, notes, and written materials. By tagging an uploaded file within a prompt, you can tell UvA AI Chat to focus on an uploaded file directly. Instead of typing out a long prompt, you can just use @documenttitle.  

For example, instead of writing an elaborate prompt to use one specific document, you simply prompt ‘use @document.docx’ and thus use less input-tokens. 

Login 

When using UvA AI Chat across devices, users can logout remotely. For example, you’ve logged into your account on the laptop of somebody else and forgot to logout. Now you’re able to login on any device and logout of the devices you want.  

Bug fixes

Various other minor issues have been resolved.