Although it remains to be seen how many of Chinese company DeepSeek's claims about its model's better performance with fewer resources can be substantiated, we had some thoughts on separating hype from the reality.

• Don’t have fear of missing out. Timnit tried to ground the public in reality on this Bloomberg News segment (beginning at minute 58)
• They’re still stochastic parrots! You might want to revisit our Stochastic Parrots Day event from March 2023.
• It's not open source. Calling DeepSeek’s model 'open source' is misleading.

A reader recently wondered if DAIR has touched upon data workers who participate in various academic research-related tasks, although they don't necessarily work on AI tasks.

The short answer is yes! In collaboration with the Weizenbaum Institute, we’re working on a project called The Ethics of Data Work (EoDW), which aims to provide best practices for academic institutions that rely on data workers, whether or not these workers are performing tasks related to AI. We can give more details on the project and its guidelines if you email us at contact@dair-institute.org.

In the latest entry in our Possible Futures series, Adrienne dreams of a future where automated tools actually help students and teachers thrive, rather than the current tools that are advertised as being able to replace teachers. You can also catch Adrienne debunking “AI for education” hype on episode 45 of Mystery AI Hype Theater 3000.