The modern world runs on software. However, there is a catch: computer code often contains programming errors--some small, some large. These glitches can lead to unexpected results--and systematic failures. "In many cases, software flaws don't make any difference. In other cases, they can cause massive problems," says Kathleen Fisher, professor and chair of the computer science department at Tufts University and a former official of the U.S. Defense Advanced Research Projects Agency (DARPA).
This refinement process may necessitate modifying D, M, and/or P. - How does the specification of unseen data relate to the specification of the data on which M was trained and tested? In traditional verification, we aim to prove property, P, a universally quantified statement: for example, for all input values of integer variable x, the program will return a positive integer; or for all execution sequences x, the system will not deadlock. So the first question for proving P of an ML model, M, is: in P, what do we quantify over? For an ML model that is to be deployed in the real world, one reasonable answer is to quantify over data distributions. But a ML model is meant to work only for certain distributions that are formed by real world phenomena, and not for arbitrary distributions. We do not want to prove a property for all data distributions. This insight on the difference in what we quantify over and what the data represents for proving a trust property for M leads to novel specification questions: - How can we specify the class of distributions over which P should hold for a given M? Consider robustness and fairness as two examples:
Nearly every time you open up a secure Google Chrome browser, a new MIT-developed cryptographic system is helping better protect your data. In a paper presented at the recent IEEE Symposium on Security and Privacy, MIT researchers detail a system that, for the first time, automatically generates optimized cryptography code that's usually written by hand. Deployed in early 2018, the system is now being widely used by Google and other tech firms. The paper now demonstrates for other researchers in the field how automated methods can be implemented to prevent human-made errors in generating cryptocode, and how key adjustments to components of the system can help achieve higher performance. To secure online communications, cryptographic protocols run complex mathematical algorithms that do some complex arithmetic on large numbers.
Why is this man smiling? He's got an army of programmers that would make some Silicon Valley CEOs green with envy. "Every year, Lloyd would give the tech division an annual speech, saying how we were a tech company," a former Goldman Sachs engineer said to Business Insider. Lloyd is Lloyd Blankfein, CEO of Wall Street investment bank Goldman Sachs. And believe it or not, Goldman has more programmers and engineers working on tech matters than Facebook.
Avalara, Inc. has acquired Indix's AI technology and expertise, with comprehensive product descriptions for more than 1 billion products sold and shipped worldwide. Avalara believes it will offer the industry's most accurate, up-to-date, and comprehensive source of global tax and product content when it combines the Indix AI expertise, technology, and content with Avalara's tax expertise and compliance content. "Global taxing authorities govern compliance through ever-changing rules and requirements, and the task of gathering and maintaining this data is core to the value Avalara provides," said Scott McFarlane, co-founder and CEO of Avalara. "We believe the combination of deep product knowledge, broad product content, and artificial intelligence technology will allow us to provide our customers the information they want and need to factor compliance into their business decision-making, and for Avalara to address more compliance requirements to support their growth." "Avalara and Indix share a similar global vision," said Sanjay Parthasarathy, CEO of Indix.