The Atticus Newsletter
November 2019
Project Updates: Phase One Complete; Partner Identified
To demystify AI, we started with a comparison test to see how AI works using our own data with our own eyes. We picked option agreements as our Phase One dataset because of the ease to anonymize.
We collected over 100 option agreements from our participants and Edgar.
We then redacted identifying information and replaced it with dummy data (i.e. replace identifying information with fake names, fake numbers and fake information).
Thanks to our valuable partners at Wachtell, Norton Rose and MoFo, we also created a separate spreadsheet summarizing the relevant clauses with answers. This is the so-called “labelled dataset.”
We then sent the anonymized option agreements and the labelled dataset to five leading Legal AI providers under NDA. Each provider had four weeks to train their AI tool using our dataset.
After four weeks, it was the demo day! We sent to each vendor 50+ brand new anonymized option agreements, and asked each to show us in real time how effective their AI tool is.
After four weeks, it was the demo day! We sent to each vendor 50+ brand new anonymized option agreements, and asked each to show us in real time how effective their AI tool is.
A few vendors also have the capability of providing answers to questions (e.g., Q: what’s the vesting cliff? A: one year).
​
With a click of a button, the system can generate a spreadsheet reporting the relevant clauses and these answers.
The Atticus participants left the demo day feeling excited. The spreadsheet that took over 100 human hours was generated within seconds. The precision score for certain questions such as “what’s the exercise price" was 96.72%. This all happened with a training set of only 100 option agreements. Imagine what is possible with a larger dataset of 200, 500, or 1,000 documents!
To complete our diligence on which partner the Atticus Project should choose, we engaged several AI scientists and data scientists to conduct confidential deeper dives with three of the five providers. After careful consideration, the Atticus Project selected eBrevia as our partner, based on its ease of use, flexibility and entrepreneurial spirit. As we are finalizing an MOU with eBrevia, you can find more details of the Atticus Project and the terms of participation in the white paper.
​
How You Can Participate
​
We are convinced that we are on a path to change the M&A diligence landscape forever. For Phase II, we are targeting commercial contracts, the most voluminous, repetitive, low-value and high-cost tasks for M&A due diligence review. We need your help to crowd source and label the dataset, and to perfect the protocols that will shape the future of our industry.
Special message to law firm associates: We need your help to label contracts. This is a great opportunity to learn about leading technologies, including AI, and their applications in the legal industry. It is also a rare opportunity to connect with clients and present your firm in the forefront of innovation.
Be a part of the Atticus movement. Email us at info@atticusprojectai.org.
Industry Trends
Wanted: Lawyers Who Understand AI
July 9, 2019 | WJS
Law firms, anticipating more cases involving artificial intelligence, are bulking up their ranks with AI experts in an effort to secure new business as the technology becomes ubiquitous across industries.
EU Report on Policy and Investment Recommendations for Trustworthy AI
June 26, 2019 | European Commission
Independent High-Level Expert Group on AI (AI HLEG) presents their ethics guidelines for trustworthy artificial intelligence. AI HLEG is an independent expert group set up by the European Commission in June 2018 and this is the second set of guidelines (first published on April 8, 2019).
Forty-Two Countries Adopt Principles for Artificial Intelligence
May 22, 2019 | Canada’s National Observer
The guidelines were drafted through an initiative led by the Organization for Economic Cooperation and Development (OECD). They aim to provide a stable environment for the development and implementation of AI through shared fundamental values.
Data Exchange Framework for Smart Cities Completes First Phase
April 16, 2019 | SmartCitiesWorld
“Data sharing is critical to solving problems that naturally extend beyond municipal boundaries,” said Nick Maynard, chief strategy officer of US Ignite.
​
Stay in Front & Additional Resources
AI & Big Data Expo North America
November 13–14 | Santa Clara Convention Center
This expo brings two days of top-level content and discussion across 5-collocated events covering, AI, big data, IoT, cyber security, cloud, blockchain, and 5G.