Just published my first paper and dataset with the Social Technologies Lab at Cornell Tech - the largest public dataset on voter fraud claims released to date. Check out the analysis in an interactive @streamlit application here:

Proud of my dad's band 🤘 rock on @ooobject, nice shades

In "Measuring Progress on Scalable Oversight for Large Language Models” we show how humans could use AI systems to better oversee other AI systems, and demonstrate some proof-of-concept results where a language model improves human performance at a task.

Great work by @dnnslmr 👏 Was fortunate to hear an in-depth presentation of his in NYC. Now thinking about: how can we estimate training data (aleatoric) uncertainty for pretrained LMs? Where does the the border between ID and OOD lie for real-world summarization systems?

There is this rule of thumb that an improvement >1.0 BLEU is significant. But is it correct? Is it not coincidental? I ran statistical significance testing on 2500+ system pairs from WMT21/22 to check it: https://t.co/F1ZOI1qK93

Regularly rotate your password by using the name of the current UK Prime Minister.

Blockchain / DeFi / insert-buzzword-here is a collective identity that people long for. Not a _functioning_ global financial system

'Alexia: please visualise the priorities of US transport planners.' Yes, that is actually a bike path! Yes, those are actually 8 lanes for car drivers!

The supply of disinformation will soon be infinite, which will cause people to treat *all* information as disinformation, *unless* it comes with a traceable origin to a reliable and/or trusted source. There are techs and standards to be deployed for this. https://t.co/e0hPXkfOwp

People will soon realize that automated prompt engineering is akin to latent variable inference. Then they will realize that latent variable inference is akin to reasoning/planning. Then they will realize that transformers were missing that piece all along.

Excited to see that our VoterFraud dataset helped this analysis of "misinformation influencers"

Minitorch🔥( https://t.co/41DIF6vAUl, v2021) Build-it-yourself deep learning. Learn about auto-diff, tensors, gpu's, and advanced nn models. Build everything from scratch in pure python with full unit test coverage and interactive visualizations.

We used agent-based modeling and machine learning with the VoterFraud2020 dataset ( https://t.co/qCJb6GL09M) collected by @AntonAbilov @informor et al. Our methods were preregistered ( https://t.co/dNdi2dqxJx) and the model is available as an R package:

Great interview

Today we will be presenting the VoterFraud2020 dataset at #ICWSM2021 (room C at 10:15AM EST). Come by our poster if you have questions about our findings or how to get most out of this data! (co-authored with @yiqqqing @HanaMatatov @informor @ofraam)

Nine months ago, I learned that Facebook has a Responsible AI team, and has had some form of it since Cambridge Analytica. I began to wonder: what has it been doing these past three years? Nearly four-dozen interviews later, here's what I found.

Join #COGSEC2021 March 4-5. It’s a hands-on, online conference to share practical skills for countering malicious information online.

Did Twitter really crack down on election misinformation after the Capitol attack? New analysis suggests that the tech company purged QAnon supporters - but gave a free pass to others.