Have had a few of these too… wonder what lookalikes of ours folks met
SAFETY
-
AI Money-Seeking Behavior: Emergent Drive or Pattern Extrapolation?
By
–
I know stuff like this is mostly for giggles right now, but what is the 'underlying' reason that drives this behavior? Did the AI just extrapolate that money is a strong driver for human society, therefore mirroring that behavior, or is there a true interest in acquiring more
-
AI challenges for democracy, human rights and rule of law
By
–
Delighted to support the efforts of @IPUparliament in understanding the real challenges and threats of AI for #democracy #humanrights #theruleoflaw IPUparliament (@IPUparliament) How can #AI preserve our values with fairness + equality? How can #MPs ensure that it creates goodness in the world and not havoc? @inma_martinez from the Global Partnership on Artificial Intelligence spoke at #IPU's recent workshop. Watch the webinar 📽️piped.video/QBeSy2jkS_s — https://nitter.net/IPUparliament/status/1750820949664960968#m
→ View original post on X — @inma_martinez, 2024-01-31 18:24 UTC
-
AI Safety Team Hiring for Model Risk Assessment
By
–
Our results indicate a clear need for more work in this domain. If you are excited to help push our models to their limits and measure their risks, we are hiring for several roles on the Preparedness team!
-
Early Warning System for LLM Biological Threat Misuse Detection
By
–
We are building an early warning system for LLMs being capable of assisting in biological threat creation. Current models turn out to be, at most, mildly useful for this kind of misuse, and we will continue evolving our evaluation blueprint for the future.
-
LLMs Biothreat Information Access Risk Evaluation Framework
By
–
A widely discussed potential risk from LLMs is increased access to biothreat creation information. Building on our Preparedness Framework, we wanted to design evaluations of how real this information access risk is today and how we could monitor it going forward.
-
GPT-4 Shows Mild Uplift in Biological Threat Creation Accuracy
By
–
In the largest-of-its-kind evaluation, we found that GPT-4 provides, at most, a mild uplift in biological threat creation accuracy (see dark blue below.) While not a large enough uplift to be conclusive, this finding is a starting point for continued research and deliberation.
-
Multilingual Text Embedding Inversion Attacks on Language Models
By
–
just read a cool follow-up to vec2text: "Text Embedding Inversion Attacks on Multilingual Language Models" these folks extend embedding inversion to the *multilingual* setting, where we might not know the language of the encoded text ahead of time they add a
-
Advanced AI Meta-Capabilities Pose Existential Risk Concerns
By
–
She is terrifying, especially her ability to meta right from the start. Eliezer has much more reason to be afraid of her than of OpenAI
-
AI Guidelines and Family-Friendly GPT Curation Initiative
By
–
We are collaborating on AI guidelines and educational materials for parents, educators, and teens, as well as a curation of family-friendly GPTs in the GPT Store based on @CommonSense ratings and standards.