When we talk about the dangers posed by artificial intelligence, the emphasis is usually on the unintended side effects. We worry that we might accidentally create a super-intelligent AI and forget to program it with a conscience; or that we'll deploy criminal sentencing algorithms that have soaked up the racist biases of their training data. But this is just half the story. What about the people who actively want to use AI for immoral, criminal, or malicious purposes? Aren't they more likely to cause trouble -- and sooner?

Machine learning models based on neural networks and deep learning are being rapidly adopted for many purposes. What those models learn, and what they may share, is a significant concern when the training data may contain secrets and the models are public -- e.g., when a model helps users compose text messages using models trained on all users' messages. This paper presents exposure: a simple-to-compute metric that can be applied to any deep learning model for measuring the memorization of secrets. Using this metric, we show how to extract those secrets efficiently using black-box API access. Further, we show that unintended memorization occurs early, is not due to over-fitting, and is a persistent issue across different types of models, hyperparameters, and training strategies. We experiment with both real-world models (e.g., a state-of-the-art translation model) and datasets (e.g., the Enron email dataset, which contains users' credit card numbers) to demonstrate both the utility of measuring exposure and the ability to extract secrets. Finally, we consider many defenses, finding some ineffective (like regularization), and others to lack guarantees. However, by instantiating our own differentially-private recurrent model, we validate that by appropriately investing in the use of state-of-the-art techniques, the problem can be resolved, with high utility.

We introduce tensor field networks, which are locally equivariant to 3D rotations and translations (and invariant to permutations of points) at every layer. 3D rotation equivariance removes the need for data augmentation to identify features in arbitrary orientations. Our network uses filters built from spherical harmonics; due to the mathematical consequences of this filter choice, each layer accepts as input (and guarantees as output) scalars, vectors, and higher-order tensors, in the geometric sense of these terms. We demonstrate how tensor field networks learn to model simple physics (Newtonian gravitation and moment of inertia), classify simple 3D shapes (trained on one orientation and tested on shapes in arbitrary orientations), and, given a small organic molecule with an atom removed, replace the correct element at the correct location in space.

We are on the verge of a technological revolution that will fundamentally alter the way we live, work, and relate to one another unlike anything humankind has experienced before. The main driver for this technological revolution is Artificial Intelligence (AI). Technological change driven by AI will change not only what we do but also who we are. It will affect our identity and all the issues associated with it: our sense of privacy, our notions of ownership, our consumption patterns, the time we devote to work and leisure, and how we develop our careers, cultivate our skills, and nurture relationships. But the development and applications of artificial intelligence can also present a dystopian threat to our collective and individual well being. From SIRI to self-driving cars, artificial intelligence (AI) is progressing rapidly. While science fiction often portrays AI as robots with human-like characteristics, AI can encompass anything from Google's search algorithms to IBM's Watson to autonomous robots and weapons systems. Artificial intelligence today is often referred to as narrow AI (or weak AI), which is designed to perform a narrow task (eg:facial recognition or only internet searches or driving a car). The other kind of Artificial Intelligence is termed general AI (AGI or strong AI) which is designed to "think," and solve problems much like humans.

With job automation, driverless cars and facial recognition hacks topping last year's headlines, artificial intelligence and machine learning have evolved from futuristic sci-fi to mainstream enterprise adoption, seizing a sizable portion of VC dollars in 2017 and pushing the federal government to define and regulate the technology. Everyone -- from investors to mainstream consumers -- is speculating about AI's role in our future as this year opens a realm of possibilities for new applications. But exactly how that technology will influence the enterprise sector remains to be seen. Fiery public debates have swirled in the tech community around whether new advancements involve tactical adoptions for enterprise, like a personalized, improved retail experience, or whether these developments signal the coming of the robot apocalypse. When it comes to business forecasts, though, I'm bullish about the potential that AI has for enterprise.

Advances in artificial intelligence have the potential to supercharge medical research and better detect diseases, but it could also amplify the actions of bad actors. That's according to a report released this week by a team of academics and researchers from Oxford University, Cambridge University, Stanford University, the Electronic Frontier Foundation, artificial intelligence research group OpenAI, and others institutes. The report's authors aren't concerned with sci-fi doomsday scenarios like robots taking over the world, such as in Terminator, but more practical concerns. Criminals, for instance, could use machine learning technologies to further automate hacking attempts, putting more pressure on already beleaguered corporate security officers to ensure their computer systems are safe. The goal of the report is not to dissuade companies, researchers, or the public from AI, but to highlight the most realistic concerns so people can better prepare and possibly prevent future cyber attacks or other problems related to AI.

As America looks for answers in the wake of the shooting massacre of 17 students at Marjory Stoneman Douglas High School in Parkland, Florida, some politicians have returned to the 1990s tactic of blaming video games for violence. Kentucky governor Matt Bevin started the show a couple of days after the shooting, and on Wednesday, Rhode Island state representative Bobby Nardolillo took it a step further. Nardolillo proposed legislation that would put a 10 percent tax on video games with an ESRB rating of Mature or higher, Rolling Stone reported. That tax revenue would be used to fund "counseling, mental health programs and other conflict resolution activities" in schools, according to the press release on Nardolillo's Facebook page. Both Nardolillo and Bevin have high ratings from the NRA. It is not controversial to say that many of the most popular video games are violent; five of the top 10 selling games of 2017 were about shooting guns and three of them would be taxed under Nardolillo's proposed regulations.

"At the time, it felt like we were in a car careening out of control and it wasn't just that everyone was saying, 'we'll be fine' -- it's that they didn't even see the car," he said. Ovadya saw early what many -- including lawmakers, journalists, and Big Tech CEOs -- wouldn't grasp until months later: Our platformed and algorithmically optimized world is vulnerable -- to propaganda, to misinformation, to dark targeted advertising from foreign governments -- so much so that it threatens to undermine a cornerstone of human discourse: the credibility of fact. But it's what he sees coming next that will really scare the shit out of you. "Alarmism can be good -- you should be alarmist about this stuff," Ovadya said one January afternoon before calmly outlining a deeply unsettling projection about the next two decades of fake news, artificial intelligence–assisted misinformation campaigns, and propaganda. "We are so screwed it's beyond what most of us can imagine," he said.

Not (yet!) of a sentient digital entity that could turn rogue and cause the end of mankind, but the exploitation of artificial intelligence and machine learning for nefarious goals. According to a group of 26 experts from various universities, civil society organizations, and think-tanks, the threat landscape can undergo dramatic changes in the next five to ten years. "The costs of attacks may be lowered by the scalable use of AI systems to complete tasks that would ordinarily require human labor, intelligence, and expertise. A natural effect would be to expand the set of actors who can carry out particular attacks, the rate at which they can carry out these attacks, and the set of potential targets," they noted in a recently released report. "New attacks may arise through the use of AI systems to complete tasks that would be otherwise impractical for humans. In addition, malicious actors may exploit the vulnerabilities of AI systems deployed by defenders."

Diversified software company Adobe Systems Inc. has said it will set up an artificial intelligence (AI) centre in Hyderabad, according to a media report. The announcement came during a meeting of Adobe CEO Shantanu Narayen and Telangana information technology minister Kalvakuntla Taraka Rama Rao, according to a report in The Economic Times. The meeting was organised on the sidelines of a three-day Nasscom India leadership forum that started this week. The event is also playing host to the World Congress on Information Technology being held in India for the first time. "As a global leader providing content creation and enterprise experience software solutions, driving innovative products is the core essence of Adobe. Hyderabad has an abundance of tech talent and a pro-business stance," Narayen was quoted as saying in the report.