Category: Transparency
-
No Deepfakes For Democracy
The main way that all of us encounter digital media — real and synthetic — is through social platforms like Facebook, Instagram, X, and TikTok. These platforms have a responsibility to protect users from deception in the form of deepfake videos, images, and audio. As the US election draws closer, these platforms must do more…
-
Join “Red Teaming In Public”
“Red Teaming in Public” is a project, originally started by Nathan Labenz and Pablo Eder in June 2024. The goal is to catalyze a shift toward higher standards for AI developers. Labenz shared the following details in the project’s announcement on X: For context, we are pro-technology “AI Scouts” who believe in the immense potential…
-
Sign the Cognition petition
The Midas Project has written a petition to Cognition, asking them to adopt an industry standard risk-evaluation policy. There are two places you can sign our petition — on our website directly, and on Change.org. (Don’t worry about signing on both — we will be combining the signature totals and cross-checking to remove duplicates before…
-
Tweet at Cognition
Despite outreach from The Midas Project, Cognition has failed to provide any information about when they will release a comprehensive risk-evaluation policy — or if they will at all. Perhaps they think they can get away with this on account of the fact that they’re still relatively under-the-radar. It’s time for the public to demand…
-
Email Cognition
Despite outreach from The Midas Project, Cognition has failed to provide any information about when they will release a comprehensive risk-evaluation policy — or if they will at all. Perhaps they think they can get away with this on account of the fact that they’re still relatively under-the-radar. It’s time for the public to demand…
-
OpenAI’s Model Spec Feedback
On May 8th, OpenAI released its Model Spec, which is a sort of charter that describes how its AI models should behave, prioritize between different values, and adhere to rules for safety, and legality. They’ve opened a form to proactively solicit feedback from the public on this document. The Midas Project is glad to see…
-
Sign Model Transparency Petition
Over the past year, OpenAI has been secretly changing ChatGPT. Researchers from Stanford and UC Berkley found that the behavior of ChatGPT changed significantly between March and June 2023, despite ChatGPT being marketed as the exact same product to customers. Source: https://arxiv.org/abs/2307.09009 In some cases, this has resulted in the model performing worse. At the…