Check out this new template that redacts information before processing it with the LLM!
SAFETY
-
AI Image Models Encode Dangerous Demographic Stereotypes
By
–
New policy brief: Millions of images are generated each day using text-to-image AI systems. Our latest brief examines how major image generation models encode a wide range of dangerous biases about demographic groups. Read or download here: https://
hai.stanford.edu/policy-brief-d
emographic-stereotypes
… -
Medical Device Regulation: Research vs Product Requirements
By
–
That's presumably because you need regulatory approval before releasing a medical device for clinical purposes.
This is a research system, not a product.
That's why regulating products is ok, but regulating research is horribly counterproductive. -
AI Alignment: White Boxes Versus Black Box Biology
By
–
AI systems are white boxes that we design As such, they are way easier to align than humans and animals which are black-boxes with all kinds of hard-wired objectives that are immutable and may or may not be "aligned", whatever you mean by that.
-
McAfee Addresses GenAI Risks and Consumer Protection Strategies
By
–
I spoke w/ McAfee's CTO about the evolution of AI-related risks, ways scammers are using GenAI & then tools McAfee is making to protect consumers. Here's a @Digiday Q&A from our interview along w/ new survey data about why AI is worrying holiday shoppers.
-
Safe AI Chatbots and RAG for Enterprise Use Cases
By
–
Understand the considerations for building safe, secure AI chatbots that stay on task and see how organizations are accelerating key use cases using retrieval-augmented generation (#RAG). Register for the webinar. https://
nvda.ws/3uHfZLC -
Hoffman-Yee Grants Fund AI Research in Foundation Models and Healthcare
By
–
New Hoffman-Yee Grants support AI research in foundation models, health care algorithms, social values in social media, and improved chip technology.
-
Self-Destructing Foundation Models: Security Against Adversarial Customization
By
–
What is a self-destructing foundation model? One that adversaries can’t customize for harmful purposes:
-
Accelerating AI Safely: Beyond the False Dichotomy
By
–
The accelerationist vs. safety dichotomy is starting to get absurd. Safety people aren’t doomers and e-accs aren’t libertarian loons. Realists are both. We must accelerate safely.
-
Billionaire Influence and AI Doomerism: Financial Corruption Concerns
By
–
You were talking about corruption?
Kersti Kaljulaid, the FLI, and yourself are under the (financial) influence of Estonian billionaire Jaan Talinn. I don't doubt that Jaan has good intentions (perhaps with a slightly excessive savior syndrome). But his and your AI doomerism is