A research tool for secure machine learning in PyTorch

Continuing the "Executive Order Series"

Understanding and Controlling the Inner Workings of Neural Networks

Geoffrey Hinton wants Musk to be banished form the Royal Society, but note for alleged "Nazi salutes".

Cloud “container” defenses have inconsistencies that can give attackers too much access. A new company, Edera, is taking on that challenge and the problem of the male-dominated startup world.

Inspired by: "ENDING ILLEGAL DISCRIMINATION AND RESTORING MERIT-BASED OPPORTUNITY"

We investigate the emergence of intuitive physics understanding in general-purpose deep neural network models trained to predict masked regions in natural videos. Leveraging the violation-of-expectati...

Meta FAIR PARTNR Benchmark

When sexual assault data, research, and women's health guidelines are erased.

This report outlines the safety work carried out for the OpenAI o3-mini model, including safety evaluations, external red teaming, and Preparedness Framework evaluations.

Rachel Chalmers and Emily Mackevicius unpack intelligence in all its forms—from songbird learning to group cognition in subway rats and humans.

Continuation of this style's series

A Year to Appreciate the Fundamentals and Authenticity

System 2 Reasoning At Scale Workshop @ NeurIPS ---

Get tickets to Open Data Science Forum, taking place 12/09/2024 to 12/11/2024. RingCentral Events is your source for engaging events and experiences.

Calculating Memory Requirements for Effective LLM Deployment

As AI systems become increasingly powerful, the need for safe AI has become more pressing. Humans are an attractive model for AI safety: as the only known agents capable of general intelligence, they ...

Generating unlimited diverse training environments for future general agents

South Korea’s National Assembly voted to lift an emergency declaration of martial law just hours after President Yoon Suk Yeol announced it. Thousands of protesters took to the streets near the Nation...

The Trouble I'm in

A well-crafted prompt is essential for obtaining accurate and relevant outputs from LLMs (Large Language Models). Prompt design enables users new to machine learning to control model output with minimal overhead. To facilitate rapid iteration and experimentation of LLMs at Uber, there was a need for centralization to seamlessly construct prompt templates, manage them, and execute them against various underlying LLMs to take advantage of LLM support tasks. To meet these needs, we built a prompt engineering toolkit that offers standard strategies that encourage prompt engineers to develop well-crafted prompt templates. It also provides functionality to enrich these templates with context obtained from RAG (Retrieval-Augmented Generation) and runtime feature datasets.

Humiliated by a Nazi officer as a teenager, she joined the French Resistance. By the time she was 20, she had killed a German soldier, survived torture and captured a supply train.

Assessing AI Benchmarks, Uncovering Issues, and Establishing Best Practices

System 2 Reasoning At Scale Workshop @ NeurIPS ---

It's in Our Faces