The issue is GPT can't safely emit long OOD sequences like data URIs during inference; it's more reliable if you ask for the result as a downloadable file.
SAFETY
-
Yolov7: Trail of Bits Reveals Model Vulnerabilities
By
–
[#Article] Trail of Bits' security assessment of #Yolov7 reveals vulnerabilities in the vision model https://actuia.com/actualite/levaluation-de-securite-de-yolov7-par-trail-of-bits-revele-les-vulnerabilites-du-modele-de-vision/
… -
One year of ChatGPT: tips and warnings in video
By
–
#ChatGPT is already 1 year old! I've used it every single day and I share everything I've learned in this video → https://youtu.be/RRptx4uIMHQ Tips but also warnings
-
AI Security and Privacy Awareness Video Impact on Loved Ones
By
–
This #AI security/privacy awareness video is a game-changer! Imagine the impact on your loved ones. #ArtificialIntelligence has incredible potential, but the risks are real @Xbond49 #AISafety #VC #data #web3 #dataprivacy #GDPR #ChatGPT #chatbots #chatgpt4 #GenerativeAI
-
Society Achieves Beneficial Artificial General Intelligence Milestone
By
–
We have achieved societally-beneficial AGI.
-
Generic AI Claims Raise Concerns About Investigation Need
By
–
Just the claims seem very generic. I could have written something better with 0 internal knowledge. I agree tho with Elon it should be investigated. Something is off with the whole thing.
-
Responsible, Safe and Reliable AI Implementation Challenges
By
–
Further, this has to be done responsibly, safely and reliably. This could be our biggest challenge.
-
Inferring LM Prompts from Output Probabilities Without API Access
By
–
haven't posted anything about (2) yet but here's the TLDR: • given LM output probabilities, we can infer what the input prompt was
• we built a model that can do this
• most APIs don't give you probabilities, but we came up with a clever algorithm to get them using logit bias -
AI Self-Destruct Mechanisms and Legal Liability Protection
By
–
Yeah like self destruct with plausible deniability for legal protection