Interpreting deep learning and machine learning models is not just another regulatory burden to be overcome. Scientists, physicians, researchers, and analyst that use these technologies for their important work have the right to trust and understand their models and the answers they generate. This talk is an overview of several techniques for interpreting deep learning and machine learning models and telling stories from their results. Speaker: Patrick Hall is a Data Scientist and Product Engineer at H2O.ai. Prior to joining H2O, Patrick spent many years as a Senior Data Scientist SAS and has worked with many Fortune 500 companies on their data science and machine learning problems.

In the final months of the Obama White House, I was honored to be invited by the President's National Economic Council to discuss the recent report, Artificial Intelligence, Automation, and the Economy. I was joined by several other venture capitalists and entrepreneurs to comment on how the tech community sees AI -- its potential for positive impact as well as the implications for our workforce. As a VC at NextWorld Capital, a big area of my investment focus is on the companies that are digitizing and automating the physical world, including drones, the Internet of Things, and artificial intelligence. As an undergraduate and graduate student at MIT studying AI in the late 90's, I saw the commercial potential of technologies such as computer vision and robotics, but now I am convinced that AI is ready to drive systemic changes to businesses and services of all kinds. This new wave of technology will have an outsized impact on what I call the "field office," operated by deskless workers are building and servicing physical goods.

We consider the problem of learning a policy for a Markov decision process consistent with data captured on the state-actions pairs followed by the policy. We assume that the policy belongs to a class of parameterized policies which are defined using features associated with the state-action pairs. The features are known a priori, however, only an unknown subset of them could be relevant. The policy parameters that correspond to an observed target policy are recovered using $\ell_1$-regularized logistic regression that best fits the observed state-action samples. We establish bounds on the difference between the average reward of the estimated and the original policy (regret) in terms of the generalization error and the ergodic coefficient of the underlying Markov chain. To that end, we combine sample complexity theory and sensitivity analysis of the stationary distribution of Markov chains. Our analysis suggests that to achieve regret within order $O(\sqrt{\epsilon})$, it suffices to use training sample size on the order of $\Omega(\log n \cdot poly(1/\epsilon))$, where $n$ is the number of the features. We demonstrate the effectiveness of our method on a synthetic robot navigation example.

Artificial intelligence can now outperform humans on a standard intelligence test. A new computational model scores within the 75th percentile, better than the average person, on a test known as Raven's Progressive Matrices. Researchers say this demonstrates that it can take on abstract visual reasoning tasks, and is a major step toward AI that can see and understand the world the way we do. Using Raven's Progressive Matrices, a nonverbal standardized test that measures abstract reasoning, the team found that their model is not only on par with humans, but performs better than many. In this example, participants choose which shape should come next in the sequence.

Soon enough, it might not be people behind the development of advanced machine learning and artificial intelligence tech, but other AI. MIT looks at the most recent work done by a range of different organizations, including Google Brain, who are working on AI that can develop machine learning software – and finds that in many cases, the results that come from machines coding other machines match or even exceed equivalent work done by humans. Does that mean even machine learning programmers are facing employment extinction? Not exactly, and not yet – efforts to create machine learning programs that best their human-designed equivalent require a lot of computing firepower thrown at the problem; Google Brain's person-besting experiment in building image recognition systems via AI development used 800 ugh-powered graphics processors working together, which is a costly endeavor to be sure. But the advantages are clear, and there's a path towards lessening the resource burden in creating these systems, too.

A Northwestern University team developed a new computational model that performs at human levels on a standard intelligence test. This work is an important step toward making artificial intelligence systems that see and understand the world as humans do. "The model performs in the 75th percentile for American adults, making it better than average," said Northwestern Engineering's Ken Forbus. "The problems that are hard for people are also hard for the model, providing additional evidence that its operation is capturing some important properties of human cognition." The new computational model is built on CogSketch, an artificial intelligence platform previously developed in Forbus' laboratory.

Ben Rossi writes in Information Age about the emergence of educational technology, and how colleges and universities could become more effective with Integrated Learning Systems by using them for more than regurgitating old styles of instruction on new equipment. He writes that ed tech should be more than just an innovative way of educational delivery, but part of the education itself by allowing students and teachers to create their own questions, answers and theories on a variety of elements on a given subject. Technology has the capacity for education to replace the currency of grades and test scores with imagination and creation in action, a necessity for an industry which spent more than $6 billion on teaching technology in 2015. Several colleges and universities are working to reform higher education into spaces for innovation and commercial development. The University of Connecticut, Arizona State University, and Princeton University are among a handful of schools encouraging students to find entrepreneurial niches and to take learning and career passions beyond the classroom.

In the October 1950 issue of the British quarterly Mind, Alan Turing published a 28-page paper titled "Computing Machinery and Intelligence." It was recognized almost instantly as a landmark. In 1956, less than six years after its publication in a small periodical read almost exclusively by academic philosophers, it was reprinted in The World of Mathematics, an anthology of writings on the classic problems and themes of mathematics and logic, most of them written by the greatest mathematicians and logicians of all time. It has influenced a wide range of intellectual disciplines -- artificial intelligence (AI), robotics, epistemology, philosophy of mind -- and helped shape public understanding, such as it is, of the limits and possibilities of non-human, man-made, artificial "intelligence." Turing's paper claimed that suitably programmed digital computers would be generally accepted as thinking by around the year 2000, achieving that status by successfully responding to human questions in a human-like way. In preparing his readers to accept this idea, he explained what a digital computer is, presenting it as a special case of the "discrete state machine"; he offered a capsule explanation of what "programming" such a machine means; and he refuted -- at least to his own satisfaction -- nine arguments against his thesis that such a machine could be said to think. But these sections of his paper are not what has made it so historically significant. The part that has seized our imagination, to the point where thousands who have never seen the paper nevertheless clearly remember it, is Turing's proposed test for determining whether a computer is thinking -- an experiment he calls the Imitation Game, but which is now known as the Turing Test. The Test calls for an interrogator to question a hidden entity, which is either a computer or another human being. The questioner must then decide, based solely on the hidden entity's answers, whether he had been interrogating a man or a machine. If the interrogator cannot distinguish computers from humans any better than he can distinguish, say, men from women by the same means of interrogation, then we have no good reason to deny that the computer that deceived him was thinking. And the only way a computer could imitate a human being that successfully, Turing implies, would be to actually think like a human being. Turing's thought experiment was simple and powerful, but problematic from the start. Turing does not argue for the premise that the ability to convince an unspecified number of observers, of unspecified qualifications, for some unspecified length of time, and on an unspecified number of occasions, would justify the conclusion that the computer was thinking -- he simply asserts it.

Donald Michie was born in Rangoon on November 11 1923, the son of James Michie and the former Marjorie Crain. From Rugby he won a classical scholarship to Balliol, becoming - according to wartime colleagues - "curator of the Balliol Book of Bawdy Verse". In 1942 he was recruited to Bletchley Park. He was put into Hut F, working to crack the Wehrmacht's "Tunny" machine, which encoded material more sensitive than that carried by the now celebrated "Enigma". The team's success gave the Allies access for the first time to German army situation reports in the run-up to D-Day, with invaluable insights into troop dispositions in France.

Quantum develops intelligent adaptive learning and assessment software solutions that inspire students to achieve more and empower instructors with real-time, actionable assessment tools to improve student learning and test performance. Accessed over the Internet anytime, anywhere, Quantum is scientifically proven to increase test scores by at least a full letter grade, accelerate learning and outperform other learning tools.