On Geoff moving due to rejection of weapons research (a mention, I'd love a more fulsome account if anyone knows one)
SAFETY
-
Why LLMs Are More Persuasive Than Humans Study
By
–
We know that LLMs are more persuasive than most humans, this study offers some tentative reasons why that may be true: LLMs produce arguments that are MORE morallly charged than humans do, and which require more cognitive work from humans to understand https://
arxiv.org/pdf/2404.09329
.pdf
… -
AI Security Risk: Prompt Injection in Context-Aware Systems
By
–
"I have access to your email, I have access to your Slack messages — when you get a message for which the context of your past can help answer it, I can just give you that draft." Hope they're thinking hard about prompt injection!
-
160 Scientists for Responsible AI in Protein Design
By
–
[#Article] Over 160 scientists commit to responsible use of AI in protein design https://actuia.com/actualite/plus-de-160-scientifiques-sengagent-pour-une-utilisation-responsable-de-lia-dans-la-conception-des-proteines/
… #AI #artificialintelligence -
AI Inbreeding Risk Threatens Quality and Diversity in Generative Models
By
–
#GenerativeAI risks "inbreeding," where AI-generated content used for training #future #AIs leads to a decline in quality and #diversity, akin to the effects of genetic inbreeding. This could degrade AI's ability to simulate human language and #creativity
-
Detecting AI-Generated Images and Model Collapse Risks
By
–
I can believe that – detecting images generated by their own image generator feels like a more tractable problem to me than detecting text from their LLMs Also model collapse for images feels more likely to me than for text, though I'm not sure I could explain that intuition
-
Curse of Recursion: 2024 Updates and Vendor Mitigation Strategies
By
–
That's about the "Curse of Recursion" from a year ago – I'm looking for a 2024 update on that. Are there new developments that counter the claims from that paper? What are the big model vendors doing (if anything) to mitigate that risk?
-
Are LLMs Really Seeds of Their Own Destruction?
By
–
Right, lots of people really want to believe that LLMs are the inevitable seeds of their own self-destruction – it's a very tempting narrative! I'm trying to understand if it's actually playing out that way
-
Artificial Data Risk at Pre-training Stage in Models
By
–
That seems like a slightly different aspect of this to me – we have seen tons of examples now of models being fine-tuned on carefully created artificial data, but to me that doesn't speak to the risk of unintentional artificial data affecting the models at the pre-training stage
-
How large model developers approach AI risk mitigation
By
–
I see lots of people outside of those organizations talking about this as a problem Presumably the people training the largest models have been thinking pretty hard about how much of a risk this is and what mitigations they can put in place What are their thoughts?