Apparently neither OpenAI's new Chat API,nor Anthropic's API for Claude allow users to request the log probabilities assigned to each token 🫤 This means they can only be used to generate text, not evaluate the probability of text under the model. (h/t Alex Lew)

Even twisting the narrative you don't have a leg to stand on Department of Health and Social Care. The inflated figure you quote £30/hr are doctors with >10yrs experience, working excessive out of hours, the most senior medic in the building at 3am. Let's compare to elsewhere 1/3

"crowd of psychometricians celebrating the demise of the MBTI personality test"

Our team at Google Brain in Paris is looking for a PhD student to work with us at the interface of differentiable programming and NLP. This is a 14-week position onsite. Contact me by email if interested!

This is absolutely disgusting

Are Emergent Abilities of Large Language Models a Mirage? Presents an alternative explanation for emergent abilities: one can choose a metric which leads to the inference of an emergent ability or another metric which does not. arxiv.org/abs/2304.15004

So proud of my wonderful friend and the most brilliant person I know, Bailey Weatherbee, for winning the Bill Gates Sr. Award!!

One of the (many) troubling things about the SCOTUS decision is that when Blum et al tried this exact same thing with Abigail Fisher it failed. But pit the model minority myth against realities of education inequity and race disparity and they get further. Divide us and conquer.

There are so many issues with anti-asian discrimination and monolithing. But using this as a bludgeon to strike down education access and opportunity equity for other Black and Brown students is insidious. And yet has been a tactic used in America for decades.

Google is hosting the first "Machine Unlearning" challenge. Yes you heard it right - it's the art of forgetting, an emergent research field. GPT-4 lobotomy is a type of machine unlearning. OpenAI tried for months to remove abilities it deems unethical or harmful, sometimes

I’m excited to share this brilliant documentary from the BBC World Service, where I was interviewed by Ellie House on AI recommendation algorithms and their impact on LGBTQ people :) Out today! bbc.co.uk/programmes/w3c…

A new APA report outlines the many obstacles faced by psychology faculty of color when pursuing tenure or promotion—including financial constraints, workplace complexities, and "inequity taxes" related to their engagement in social justice issues. apa.org/news/press/rel…

Very excited that our DICES - #Diversity in #ConversationalAI #Safety #Evaluation paper arxiv.org/abs/2306.11247 got accepted at the NeurIPS Conference Datasets and Benchmarks Track with Vinodkumar Prabhakaran Alicia Parrish [email protected] Mark Diaz, Christopher Homan, Greg Serapio-Garcia, Ding Wang

So excited to have this paper be at NeurIPS Conference D&B track. This is the flagship paper (a few more to follow) out of a multi-year cross-org collaboration that was led by the amazing Lora Aroyo. It adds to the recent line of papers interrogating disagreements in data annotations.

How do we probe for particular moral stances mimicked by LLMs? What are the potential risks and downstream consequences of prompting models with moral stances? Check out this cool project I worked on at Google AI with Marwa Abdulhai + Natasha Jaques!

My students and I are running a creative writing competition with a twist. You'll write three short pieces, but sometimes you'll work by yourself and sometimes with an AI. Winner gets £500 and 5 runners up get £100 each. Takes 40 mins, have a go & pls RT! tinyurl.com/camcreativewri…

Proud to see Google AI’s latest blog post feature responsible AI + rater diversity research I’ve worked on w/ Lora Aroyo, Alicia Parrish, Mark Díaz, @justaturbo, Vinodkumar Prabhakaran, [email protected], and others! #GoogleResearch