January 2022

AI2 Newsletter

The Aristo team logo and the Macaw logo

Top story

Macaw: smaller and better than GPT-3... and free!

Macaw, a general-purpose question-answering system, has been available as a free open-source model since October. This week, the Aristo team at AI2 released a brand new interactive demo of this powerful new model, allowing anyone to explore the answers it can provide as compared to the popular language model GPT-3.

Macaw can successfully tackle a broad range of question types including general knowledge, meta reasoning, hypothetical, and story understanding. Not only does Macaw outperform the GPT-3 model on our suite of 300 challenge questions, but it is also an order of magnitude smaller, making it a very promising evolution in the commoditization of powerful language models.

America Needs AI Literacy Now

A stack of colorful books
In this featured article from the pnw.ai hub, AI2's Oren Etzioni and Nicole DeCario share findings from an AI literacy survey they published in the summer of 2021. Of 1,547 US adults, only 16% of participants "passed" with a score of above 60%. To learn more or test yourself with the quiz, visit the link below.

AI2 researchers recognized with NeurIPS 2021 Outstanding Paper Award

The Neural Information Processing Systems 2021 Conference recognized a team including several researchers from AI2 with their Outstanding Paper Award for MAUVE: Measuring the Gap Between Neural Text and Human Text using Divergence Frontiers. Written by Krishna Pillutla, Swabha Swayamdipta, Rowan Zellers, John Thickstun, Sean Welleck, Yejin Choi, and Zaid Harchaoui, this paper presents a model for comparing the distribution of model-generated text with the distribution of human-generated text. Out of 9,122 full paper submissions to NeurIPS, only six were distinguished with the Outstanding Paper Award. The paper was presented at the NeurIPS Conference on December 7th.

S2AG: the new Semantic Scholar Academic Graph for developers

Now that the popular Microsoft Academic Graph is no longer available, Semantic Scholar has made its corpus of 200 million papers and collection of APIs a viable alternative for those looking to build scholarly applications. Providing these resources to the global research community helps AI2 scale progress and innovation in science.

Want to learn more about AI2's work or connect with an expert? Reach us at press@allenai.org.

A portrait of Nayha Auradkar

Scholarship Recipient Nayha Auradkar

UW Allen School senior and rising Master’s student Nayha Auradkar was named the recipient of AI2's 2022 Allen AI Outstanding Engineer Scholarship. Among an excellent cohort of applicants, Nayha stood out with her strong motivation for using technology to make positive impacts for people with disabilities.

Learn more ➞
Coast watch team members help the lost skipper captain onto their boat

Skylight helps locate a missing skipper

When Peter Ming and his sailing yacht "OUMA" missed their expected arrival date to Palawan, the National Coast Watch Center of the Philippines employed their best technology — including AI2's platform Skylight — to successfully locate the missing skipper after 15 days at sea. That this unexpected use of the Skylight platform helped in this critical rescue mission indicates the incredible potential of this technology.

More from us

➞ What Does It Mean for AI to Understand?
➞ Can a Moral Machine Make Good Decisions about Animals?
Semantic Scholar celebrates 200 million archived papers
➞ Doug Downey/Semantic Scholar: Applying Cutting Edge NLP at scale
Visit us: allenai.org | Follow us: @allen_ai | Work with us!

AI2 Color Logo