This is what happens when you release a poorly tested #LLM in the hands of real people who now suffer the consequences @OfficeforAI @SciTechgovuk @theCAIDP @GPAI_PMIA
SAFETY
-
ChatGPT security breach exposes user conversation history titles
By
–
we had a significant issue in ChatGPT due to a bug in an open source library, for which a fix has now been released and we have just finished validating. a small percentage of users were able to see the titles of other users’ conversation history. we feel awful about this.
-
ARC Partnership Advances AI Safety and System Interpretability
By
–
Working with ARC is part of our overall vision for AI safety and evaluation, as we work to build more steerable, predictable, and interpretable systems. You can read more about our approach to safety and the societal impacts of AI systems here:
-
ARC Evals Tests Models for Deployment Readiness Assessment
By
–
We’re excited to have ARC Evals test our models to assess their readiness for deployment. We look forward to sharing more about our approach to evaluating our systems in the coming months.
-
ARC advances AI alignment evaluation and security measurement
By
–
We strongly agree there’s much more work to be done on alignment, security, and measurement. You can read about ARC’s specific approach to evaluation here: https://
evals.alignment.org/blog/2023-03-1
8-update-on-recent-evals/
… -
Understanding AI Model Anthropomorphization and Training Data
By
–
it seems like one of the inherently confusing things about these models. I know that they've learned to talk about their feelings and desires because that's in the training data, but I think it contributes to the misunderstanding.
-
Who Should Decide How AI Systems Behave?
By
–
How Should AI Systems Behave, and Who Should Decide? https://
openai.com/blog/how-shoul
d-ai-systems-behave
… @ASMEdotorg @3DSNorthAmerica @MargaretSiegien @3DSdelmia @3DStherese @Cindybolt61 @fogoros @DrFerdowsi @CRudinschi @PawlowskiMario @IIoT_World @MEngineeringMag #Science #Engineering #Technology #SET -
Black Box AI Systems: The Reproducibility and Transparency Crisis
By
–
Without knowing how these systems are built, there is no reproducibility. You can't test or develop mitigations, predict harms, or understand when and where they should not be deployed or trusted. The tools are black boxed.
-
Model Safety: Mitigation Without Full Release, Transparency Needed
By
–
There's a lot of ways to mitigate harms without having to publicly release the entire model. There are many papers on auditing, datasheets, transparency etc. With GPT3 we knew the training data. With GPT4 we don't. Without that, we're all looking at shadows in Plato's cave.
