Multiple events where people adjusted a seat to attempt to get at trapped electronics. The seats then applied substantial torque which via mechanical action punctured the seal of a lithium-ion battery, causing rapid combustion. Fire in midair is not wonderful, etc etc.
SAFETY
-
Exclusion from AI Safety Summit Second Day Raises Questions
By
–
I wasn't invited to the 2nd day of the AI Safety Summit
-
Racial Disparities in ML Medical Imaging for Dermatology
By
–
Researchers identify new forms of racial disparity in medical imaging for dermatology by analyzing ecosystems of ML models. @StanfordCRFM @RishiBommasani
-
AI Existential Risk: Assessing Extinction Probability in 30 Years
By
–
Yes. It would be crazy not to consider this possibility.
Have you seriously considered the possibility it might wipe us all out and if so what probability do you put on that happening in the next 30 years. If you are really a Bayesian you should be able to give me a probability. -
Open-Source AI as Essential Counterbalance to Proprietary Concentration
By
–
IMO your points apply to non-open-source AI as well at 100x the scale (as proprietary AI is deployed massively) and open-source is the only way to keep non-open-source in check. Without it, you'll have extreme concentration of power/knowledge/secrecy with 1,000x the risk.
-
Singapore GenAI Safety Evaluation Sandbox Initiative Unveiled
By
–
We congratulate @IMDAsg and AI Verify Foundation on their incredible initiatives in GenAI safety: unveiling a first-of-its-kind GenAI Evaluation Sandbox during @SwitchSingapore 2023!
-
Embedding Societal Values in Social Media AI Systems
By
–
What if social media AIs could embed societal values vs. only focusing on the individual experience? A commentary by Stanford scholars in @journalsafetech explores the possibility. @msbernst
-
Chain-of-Verification Reduces Hallucination in Large Language Models
By
–
Chain-of-Verification Reduces Hallucination in Large Language Models https://
arxiv.org/abs/2309.11495 @shehzaadzd Mojtaba Komeili @jingxu_ml @robertarail Xian Li @real_asli @jaseweston -
Sparse Autoencoders Reveal Interpretable Language Model Features
By
–
Sparse Autoencoders Find Highly Interpretable Features in Language Models https://
arxiv.org/abs/2309.08600 @HoagyCunningham @aidanprattewart @loganriggssmith @Robert_AIZI @leedsharkey -
Chain-of-Thought Reasoning as Policy Improvement Operator
By
–
Chain-of-Thought Reasoning is a Policy Improvement Operator https://
arxiv.org/abs/2309.08589 @hughbzhang David C. Parkes