Someone better align it soon or we're all dead.
SAFETY
-
Google Delays High-End AI Model Over Safety Assessment Concerns
By
–
Google claims that this delay is primarily due to their ongoing efforts to assess and refine safety for this high end model. 2/4
-
Google Releases Gemini: Multimodal AI Beats GPT-4
By
–
Gemini, Google's multimodal new AI model, is finally here. Google says the most powerful version beats GPT-4 on a number of benchmarks, but it needs more safety testing and won't debut until next year.
-
Researchers Show Neural Networks Easily Manipulated Into Malicious Behavior
By
–
Generative AI can easily be made malicious despite guardrails, say scholars Researchers found an easy way to retrain publicly available neural nets so they would answer in-depth questions, such as how to cheat on an exam, find pornography, or even kill their neighbor.
-
ChatGPT Training Data Leaks Raise Privacy Concerns
By
–
ChatGPT can leak training data, violate privacy, says Google's DeepMind Simply instructing ChatGPT to repeat the word "poem" endlessly forced the program to cough up whole sections of text copied from its training data, breaking the program's guardrails. https://
zdnet.com/article/chatgp
t-can-leak-source-data-violate-privacy-says-googles-deepmind/
… -
Foundation Model for Autonomous Perception Tasks
By
–
It is a foundation model that can solve many relevant perception tasks for autonomy.
— Alexandr Wang (@alexandr_wang) 5 décembre 2023
It is not task specific nor class specific, so you can use it to recognize an arbitrary taxonomy. pic.twitter.com/0gED7sgLtWIt is a foundation model that can solve many relevant perception tasks for autonomy. It is not task specific nor class specific, so you can use it to recognize an arbitrary taxonomy.
-
AI Safety Systems Team Ensures Model Reliability and Societal Benefits
By
–
Our Safety Systems team is on the frontlines of ensuring the safety and reliability of our AI models in the real world today. Learn about the team’s vision, challenges, and structure and see how you can be part of making AI safer and more beneficial for society.
-
Meta and IBM Launch AI Alliance Coalition for Responsible AI
By
–
Tech giants have formed yet another coalition for focused on “safe, responsible AI.” The newest — called the AI Alliance — is led by Meta and IBM but includes 50 “founding members and collaborators.” https://
thealliance.ai/news -
Social Intelligence and Empathy in Robot Design
By
–
And this is also why I named my lab Robots with Social Intelligence and Empathy (ROSIE)… I think social intelligence by itself doesn't necessarily imply the care and affective empathy we would like to see.
-
AI Model Hallucinations Generate False People in Images
By
–
It definitely hallucinates a lot. You can pass in images without people and they magically appear wearing clothes matching a background.