On a recent work trip, I found myself in a swanky-but-still-hip office of a private tech firm. I was drinking a freshly frothed cappuccino, eyeing a mini-fridge stocked with local beer and standing amidst a group of hoodie-clad software developers typing away diligently at their laptops against a backdrop of Star Wars and xkcd comic wallpaper. I wasn't in Silicon Valley: I was in Johannesburg, South Africa, meeting with a firm that is designing machine learning (ML) tools for a local project backed by the U.S. Agency for International Development. Around the world, tech startups are partnering with NGOs to bring machine learning and artificial intelligence to bear on problems that the international aid sector has wrestled with for decades. ML is uncovering new ways to increase crop yields for rural farmers.

Large amounts of labeled data are typically required to train deep learning models. For many real-world problems, however, acquiring additional data can be expensive or even impossible. We present semi-supervised deep kernel learning (SSDKL), a semi-supervised regression model based on minimizing predictive variance in the posterior regularization framework. SSDKL combines the hierarchical representation learning of neural networks with the probabilistic modeling capabilities of Gaussian processes. By leveraging unlabeled data, we show improvements on a diverse set of real-world regression tasks over supervised deep kernel learning and semi-supervised methods such as VAT and mean teacher adapted for regression.

I recently read an article (paywall) in the WSJ about Paul Allen's Vulcan initiative to curb illegal fishing. It's insightful and sheds light on Big Data techniques to address societal problems. After thinking on the story, it struck me that it could be used as a pedagogical tool to synthesize data science with domain knowledge. To me, this stands as the biggest limitation of what I refer to as'data science thinking'– letting technical skills drive the analysis, only later incorporating domain understanding. This post somewhat reads like a case note from business school and the idea is to get data scientists, product managers and engineers talking earlier on in the process.

Paul Steinberg, CTO of Motorola Solutions, speaks to Sam Fenwick about his company's efforts to use AI and machine learning to bring the right data to the user in the right way Paul Steinberg presides over a huge range of research and development activities, ranging from RF engineering and wireless network architectures to drones and robotics. He also manages Motorola Solutions Venture Capital's portfolio and plays a key role in managing Motorola Solutions' intellectual property. One of the things the company is moving towards is a virtual partner – a combination of AI and natural language processing, which allows someone in the field to verbally request information and give commands without talking to a human. Part of the thinking behind this is that people speak faster than they can type, and the need for field workers to stay aware of their surroundings. "The way you and I consume [mobile data] is a slab of black glass, [but the] fundamental imperative [for a police officer, etc] is eyes-up, hands-free. That slab of black glass [is] exactly the opposite: eyes-down, hands-busy. A big part of how we're navigating this problem is around ethnographics and human factors research – living a day in the life of our users and then [working] with the technologists and designers."

One day in late April or early May, Google removed the phrase "don't be evil" from its code of conduct. After 18 years as the company's motto, those three words and chunks of their accompanying corporate clauses were unceremoniously deleted from the record, save for a solitary, uncontextualized mention in the document's final sentence. Google didn't advertise this change. In fact, the code of conduct states it was last updated on April 5th. The "don't be evil" exorcism clearly took place well after that date. Google has chosen to actively distance itself from the uncontroversial, totally accepted tenet of not being evil, and it's doing so in a shady (and therefore completely fitting) way.

MajorMajor CyberCyber--AttackAttack UK Internet Service ProviderUK Internet Service Provider EstimatedEstimated $$$$$$ Loss Loss $55$55 MillionMillion 12. 1717thth Nov 2015Nov 2015: "Islamic State is Plotting: "Islamic State is Plotting Deadly CyberDeadly Cyber--Attacks":Attacks": George OsborneGeorge Osborne 12CyberVision: 2015CyberVision: 2015 -- 20252025 *** 21stC Cybersecurity Trends *** London, UK:: 15th December 2015 Dr David E. Probert: www.VAZA.com

The hack, featuring software engineers from Microsoft who had travelled from across Europe and Africa to work with OS's machine learning team, used the city of Hull as a testbed. The trained machine model finished the week by correctly identifying 87% of the roof types it was shown. In its training the model was shown 500 flat roofs and 500 hipped/gabled roofs, set a confidence limit of 90%, which means it must be 90% confident or more for its assessment to count. Isabel Sargent, Senior Research and Development Scientist at OS, says: "Thanks to the excellence of the Microsoft team we have been able to work out together how to stream this machine captured data into our database for if and when we're ready to put machine learning into production. It's already very accurate, going from zero to 87% accuracy in just one week, but we need to increase its success rate. Although much slower, humans typically have an error rate of around 5%."

A USC-led study of violent protest has found that moral rhetoric on Twitter may signal whether a protest will turn violent. The researchers also found that people are more likely to endorse violence when they moralize the issue that they are protesting--and when they believe that others in their social network moralize that issue, too. "Extreme movements can emerge through social networks," said the study's corresponding author, Morteza Dehghani, a researcher at the Brain and Creativity Institute at USC. "We have seen several examples in recent years, such as the protests in Baltimore and Charlottesville, where people's perceptions are influenced by the activity in their social networks. People identify others who share their beliefs and interpret this as consensus. In these studies, we show that this can have potentially dangerous consequences."

At least half of the world's population cannot obtain essential health services. But low-cost, easy-to-use technologies powered by Artificial Intelligence (AI) promise to deliver quality and affordable health care to the people who need it most, no matter how hard to reach. At the AI for Good Global Summit last week, entrepreneurs, AI experts, academics and UN representatives described many AI technologies for health, allowing for the early detection of various pathologies such as osteoarthritis, diabetic retinopathy, child malnutrition, snakebites and others. These technologies don't place a heavy burden on doctors, and can lead to prompt diagnosis and effective treatment. They agreed that AI can add tremendous value in developing countries where there is a low density of physicians.

Coping with harsh conditions, rather than social challenges, was chiefly responsible for boosting the size of our brains, a new study has found. The research found'ecological' challenges like finding food and lighting fires boosted the capacity of our ancestors to think ahead. The finding may settle a decades-long debate on the origins of human intelligence and our social relationships, scientists said. The human brain got so big because life was tough on the African savannah around two million years ago, according to new research. The human brain has tripled in size compared to the white matter of our ancestor Australopithecus afarensis, which roamed the Earth more than 3 million years ago.