ChatGPT overuses certain words, notably 'delve'

Why?

Turns out it was trained by RLHF workers from Africa

In Nigeria, “delve” is much more frequently used in business English than in England or the US. So we are ending up with an AI system that writes slightly like an African

🔍 Looking for the best on journalism & AI for our community!

What are your go-to newsletters, podcasts, or Twitter accounts focused on journalism and AI? Share them here and let me know why they stand out!

I’m curating top resources for our Journalists on Hugging Face…

Life of an immigrant researcher:
-Got #ICLR2024 spotlight paper acceptance
-I'm in WA and need to fly to CA to apply for a Schengen visa (no center in WA)
-Unfortunately, I couldn't secure a visa appointment.
Thought getting a spotlight paper was the toughest part!

Goat Life is a Malayalam movie about an Indian worker in Saudi Arabia, in worst conditions in the desert. While in the West labour conditions in the gulf are often discussed, rarely these include what it actually means for people affected. Watch it!
thegoatlifefilm.com

The article gives a great write-up of why it is crucial to ensure information accessibly to speakers of *all* languages.

🚨BIG STORY🚨Swiggy has gamified health insurance for delivery workers, changing insurance tiers depending on a worker's weekly performance. Varsha Bansal breaks an important story for Rest of World restofworld.org/2024/swiggy-he…

NeurIPS managed to get high school students to the conference before finding a way to enable participation in case of travel restrictions and rejected visas 🤔

NeurIPS should not be doing a high school track.

The main effect will be to give the most-privileged yet another tool with which to outcompete less-privileged kids on college admissions.

We need more paths for getting into research at *later* stages of life, not earlier ones.

Very very excited to share that Eryk Salvaggio and I's collaborative art piece, called 'Because of You' -- has been accepted to the #CVPR2024 2024 AI Art Gallery!

A short thread:

huggingface.co/spaces/evijit/…

We are hiring Masakhane The Masakhane Research Foundation, through the support of FAIR Forward, an initiative of the German Development Cooperation (GIZ), is glad to announce two Data Governance Fellowship positions. ->Open Positions masakhane.io/open-positions

Did you know that, if you evaluate the same model, with the same prompt formatting & the same fixed few-shot examples, only changing
♻️the order in which the few shot examples are added to the prompt ♻️
you get a difference of up to 3 points in evaluation score?

Today, the U.S. government is establishing new rules for its own use of AI, and they’re a HUGE deal

Buckle up folks — we’re going to unpack how they work and why they’re so good 🧵

blog.mozilla.org/netpolicy/2024…

So excited to announce an event The GenLaw Center has been working on!

We're discuss the misconceptions b/w the technical capabilities of evaluating generative AI, and what policymakers and civil society want...

April 15th Georgetown Law Tech Institute, and live on zoom:  dc-workshop.genlaw.org

For years I’ve been interviewing data annotation workers who are the lifeblood of the AI industry. For years I’ve heard the same story: the platforms they work for wield total power, leaving them precarious & vulnerable to exploitation. A horrible example of this just happened 1/

The risks of wrongly deployed AI and its bias are rarely as clear and frightening as in the context of healthcare.

Regarding risks relating to AI use, the European Parliament creating a governance board is a step in the right direction. We have discussed some of the existing mitigation strategies for misuse of AI in our paper on dual use in NLP aclanthology.org/2023.findings-…

💥 New paper 💥

We discover a form of covert racism in LLMs that is triggered by dialect features alone, with massive harms for affected groups.

For example, GPT-4 is more likely to suggest that defendants be sentenced to death when they speak African American English.

🧵

Our tutorial at #ICWSM on how to reuse #Wikipedia data in research, including how to analyse editors behaviour in a respectful manner, and data access to the information of different modalities in the Wikimedia universe (articles, open license images, structured data, …)

China has released the <Basic Security Requirements for Generative AI Services>📖setting standards for data, model safety, and security protocols, including assessment guidelines🔍🧵