Bad things happen when AI is trained on synthetic data.
Source: Model collapse explained: How synthetic training data breaks AI
Bad things happen when AI is trained on synthetic data.
Source: Model collapse explained: How synthetic training data breaks AI
ChatGPT Co-Creator, Greg Brockmanformer speaking at the University of North Dakota on AI and Higher Education on 9/26 – Streamed event.
Source: ChatGPT Co-Creator, former UND Student Greg Brockman coming to campus – UND Today
Created through a survey of 2,022 participants.
Source: Top 100 Tools for Learning 2023 – Results of the 17th Annual Survey published 4 September 2023
Preliminary guidance to federal institutions on their use of generative AI tools in Canada. Seems like the language would be easy to apply by a number institutions.
Benjamin Breen writes about teaching history with AI – hallucinations and all. The article includes the prompts that Breen used for a number of class activities including a Medieval Plague Simulator.
Source: MIT Technology Review
Open AI has published a guide for teaching with ChatGPT that includes sample prompts, an explanation of how ChatGPT works and its limitations including bias.
Source: Open AI Teaching with AI
This study by Akiba and Fraboni assessed the responses by the free version of Chat GPT to common advising questions. The authors concluded the AI gave high quality answers that were, “written in an authoritative yet supportive tone”.
SynthID is a new tool from Google for watermarking and identifying AI-generated images. While the watermark can’t be seen with the human eye, SynthID can scan an image to see if it was created with generative AI.