The ever-expanding capabilities of AI evoke fear in some quarters. On one side you have prominent AI researchers, like Stuart Russell, warning about impending disaster, and on the other, you have equally notable scholars, like Andrew Ng, offering a more tempered view. I support the notion that we are far from human-level AI, and the warning bells are not sounding just yet. One thing we know for sure is AI is a tool, not a being, and it is up to us to deploy this tool for good. AI and Covid 19 is an example of how this can happen.
[Related article: Trending Data Science Topics Coming to ODSC West 2020]
Our focus at the Allen Institute for AI (AI2) is to conduct high-impact AI research and engineering in service of the common good. Semantic Scholar, our flagship project, was founded on a simple question, “What if the cure for intractable disease lives buried within the scientific literature?” This hypothesis has been put to the test in this age of Covid. Since the beginning of the year, thousands of papers have emerged on the topic. Which is why our team leapt at the opportunity to collaborate with the White House, Google, Microsoft, Amazon, The Chan Zuckerberg Initiative, and other partners to create the Covid-19 Open Research Dataset (CORD-19): a free resource of over 275,000 scholarly articles about the novel coronavirus. Since its launch in mid-March, this machine-readable corpus has been viewed over 2M+ times and was the focus of the most successful Kaggle competition to date. Over 50 tools, ranging from question/answering search to data visualization, are now publicly available to assist researchers and the medical community in solving the Covid crisis. Much of this work was possible due to a fundamental sea-change in the world of NLP. Typically, NLP looks at local context, something like 512 tokens around a particular word. Through Longformer, this has expanded to over 32k tokens allowing for encoding of full, or even multiple, papers.
The implications of this work are far-reaching and a testament to the potential of AI to accelerate science. Not only has this work made AI accessible to a wide range of researchers, but it has also positioned AI as a key contributor in fighting disease and demonstrated how broad access to scientific literature for automated analysis and discovery can accelerate advancements in all aspects of research. We’ve also witnessed the power of collaboration as scholars, researchers, data analysts, and scientists from various disciplines have coalesced around a shared goal.
Of course, issues of privacy, fairness, and economic implications cannot be overlooked. We cannot abandon these and other core values for rapid advancement. AI is here, and it can be used for good and evil. It is up to us to decide on the use of AI and COVID 19.
Editor’s note: Oren is a speaker for ODSC West 2020. Check out his talk on AI and COVID 19, “Semantic Scholar and the Fight Against COVID-19,” there.
Dr. Oren Etzioni is Chief Executive Officer at AI2. He has been a Professor at the University of Washington’s Computer Science department since 1991. His awards include Seattle’s Geek of the Year (2013), and he has founded or co-founded several companies, including Farecast (acquired by Microsoft). He has written over 100 technical papers, as well as commentary on AI for The New York Times, Wired, and Nature. He helped to pioneer meta-search, online comparison shopping, machine reading, and Open Information Extraction.