NYU Data Science newsletter – June 21, 2016

GROUP CURATION: N/A

We study the effectiveness of neural sequence models for premise selection in automated theorem proving, one of the main bottlenecks in the formalization of mathematics. We propose a two stage approach for this task that yields good results for the premise selection task on the Mizar corpus while avoiding the hand-engineered features of existing state-of-the-art models. To our knowledge, this is the first time deep learning has been applied to theorem proving.

Google has done a great job with maps but is struggling in disrupting itself with general search. Bing has done some interesting work in structured search, but nothing revolutionary over Google. Siri had an early shot, but Apple delayed further investments for years. Facebook had some good initiative, but we have yet to see something concrete. IBM Watson on the enterprise side, or the academic Wolfram Alpha could have a good hand, but we have yet to see traction. And then there are the smaller newcomers such as Graphiq, ViV, and Hound. The race to own the future of search has started, and one of the biggest businesses in history is up for grabs.

TK: Google bundle

Despite misperceptions to the contrary, farming in the 21st century is a high-tech endeavor. We’re not just talking about genetically modified crops or biotech-derived pesticides though; farm vehicles like tractors and combines are now networked to the cloud and in many cases are even capable of driving themselves. To find out more about what the modern technofarm is all about, I drove up to Clear Meadow Farm in Harford County, Maryland to meet farmer Greg Rose and his self-driving John Deeres.

At Facebook AI Research, we’ve published a set of papers on stabilizing adversarial networks in collaboration with our partners, starting with image generators using Laplacian Adversarial Networks (LAPGAN) and Deep Convolutional Generative Adversarial Networks (DCGAN), and continuing into the more complex endeavor of video generation using Adversarial Gradient Difference Loss Predictors (AGDL). Regardless of what kinds of images or videos we gave to these systems, they would start learning and predict plausible scenarios of the world.

Multispectral scanning reveals ancient text on the fabled Antikythera Mechanism, and suggests the machine was a mechanical textbook.

Plaid Technologies, a builder of infrastructure for financial services companies, just raised $44 million led by Goldman Sachs Investment Partners, continuing the trend of Wall Street banks investing in fintech startups.

Machine learning is increasingly at the core of everything we build at Twitter. It’s powering much of the work we’re doing to make it easier to create, share, and discover the very best content so that every time you open Twitter you’re immersed in the most relevant news, stories, and events for you.

Today, we’re very excited to announce that we’re expanding our capabilities in machine learning by acquiring Magic Pony Technology, a London-based technology company that has developed novel machine learning techniques for visual processing.

Enabling a computer to understand a document so that it can answer comprehension questions is a central, yet unsolved goal of NLP. A key factor impeding its solution by machine learned systems is the limited availability of human-annotated data. Hermann et al. (2015) seek to solve this problem by creating over a million training examples by pairing CNN and Daily Mail news articles with their summarized bullet points, and show that a neural network can then be trained to give good performance on this task. In this paper, we conduct a thorough examination of this new reading comprehension task. Our primary aim is to understand what depth of language understanding is required to do well on this task. We approach this from one side by doing a careful hand-analysis of a small subset of the problems and from the other by showing that simple, carefully designed systems can obtain accuracies of 72.4% and 75.8% on these two datasets, exceeding current state-of-the-art results by over 5% and approaching what we believe is the ceiling for performance on this task.

Four foundations announced their support for ASAPbio (Accelerating Science and Publication in Biology), a scientist-driven effort with a mission to promote the use of preprints in the life sciences. The combined total provisional funding — from the Alfred P. Sloan Foundation, the Gordon and Betty Moore Foundation, the Laura and John Arnold Foundation and the Simons Foundation — is $400,000 for work to be conducted over the next 18 months.

This is a book I am writing over summer 2016. If you want to get a free draft copy of each chapter as it is finished, please sign up by Friday Jun 24th [today!] for my mailing list. … Historically, the only way to learn how to make these “strategy” decisions has been a multi-year apprenticeship in a graduate program or company. I am writing a book to help you quickly gain this skill, so that you can become better at building AI systems.

People often rank public speaking as the number one fear that they face. New cloud-based technology from researchers at the University of Rochester lets speakers polish and practice at home in front of their computer camera, while the analysis provides instant feedback about improvement.