AI2 Newsletter June 2019

Send this newsletter to friends and colleagues who want to stay up to date with research and events at the Allen Institute for Artificial Intelligence. Subscribe here .

AI2 NEWSLETTER | June 2019

Gender parity in computer science research could be 100 years away

A new analysis by AI2 of over 3 million computer science papers reveals that women will not reach parity with men in authorship of published CS research in this century, if current trends hold.

R ead the study, or learn more in press coverage of this compelling finding:

The Gender Gap in Computer Science Research Won’t Close for 100 Years — New York Times
It could take 118 years for female computer scientists to match publishing rates of male colleagues — Science Magazine
A study about studies suggests men will still prevail in computer science in 2100 — GeekWire

Grover — A State-of-the-Art Defense against Neural Fake News

AI2 researchers are addressing the threat of machine-generated disinformation with a fake news detector—and generator—called Grover .

Try it for yourself, and read more about the research and our newest findings in the AI2 Blog.

AI2 opens a new office in Tel Aviv

We're excited to announce the opening of AI2 Israel! This vibrant new Tel Aviv office including Yoav Goldberg, Jonathan Berant, Reut Tsarfaty, and Ron Yachini will will focus on bringing people closer to information with advanced language-centered AI.

Learn more at allenai.org/ai2-israel.

RECENT PRESS

To detect fake news, this AI first learned to write it — TechCrunch
Can Researchers Create Commonsense Artificial Intelligence? — NYAS
True story: Fake news generator created by neural network researchers aims to fight disinformation — GeekWire
No, this AI can’t finish your sentence — ZDNet
Allen Institute for Artificial Intelligence to Open Israeli Branch — CTech

NEW RESEARCH

COMET: Commonsense Transformers for Automatic Knowledge
Graph Construction

Antoine Bosselut, Hannah Rashkin, Maarten Sap, Chaitanya Malaviya, Asli Celikyilmaz, Yejin Choi

We introduce COMET, a new framework for automatic knowledge base construction that is frequently capable of producing novel commonsense knowledge that human evaluators deem correct. These positive results point to future work in extending the approach to a variety of other types of knowledge bases.

HellaSwag: Can a Machine Really Finish Your Sentence?

Rowan Zellers, Ari Holtzman, Yonatan Bisk, Ali Farhadi, Yejin Choi

HellaSwag is a commonsense inference challenge dataset that is trivial for humans but difficult even for the most advanced language models. This approach also suggests a new path forward for NLP research, in which benchmarks co-evolve with the evolving state-of-the-art in an adversarial way, so as to present ever-harder challenges.

The Risk of Racial Bias in Hate Speech Detection

Maarten Sap, Dallas Card, Saadia Gabriel, Yejin Choi, Noah A. Smith

We investigate how annotators’ insensitivity to differences in dialect can lead to racial bias in automatic hate speech detection models, potentially amplifying harm against minority populations.

See all of AI2's recent publications on our website →

MORE FROM AI2

Featured AI2er Matthew Peters talks about his path to NLP research, what's coming up for him this summer, and, offers some advice for aspiring researchers.

AI2 researcher Matt Gardner proposes simple steps to get us to a paper review process that is more equitable for everyone:
5 steps to reconciling pre-prints and blind review

Learn more about Asila Maksumova, this year's recipient of our annual Allen AI Outstanding Engineer Scholarship for Women and Underrepresented Minorities.

Check out computer vision researcher Ani Kembhavi's series on Visual Question Answering, new on the AI2 Blog:

• Vanilla VQA
• May I have your attention please?
• The lure of the outer project

Check out more on the AI2 Blog →

Were you forwarded this newsletter from a friend or colleague? To stay up to date with research and events at AI2, subscribe here.