More and more of our experience online is personalized. Search engines, news outlets and social media sites have become quite smart at giving us what we want. Perhaps Ali, one of the hundreds of people I've interviewed about our emotional attachment to technology, put it best: "Netflix's recommendations have become so right for me that even though I know it's an algorithm, it feels like a friend." Personalization algorithms can shape what you discover, where you focus attention, and even who you interact with online. When these algorithms work well, they can feel like a friend. At the same time, personalization doesn't feel all that personal. There can be an uncomfortable disconnect when we see an ad that doesn't match our expectations. When personalization tracks too closely to interests that we've expressed, it can seem creepy. Personalization can create a filter bubble1 by showing us more of what we've clicked on before, rather than exposing us to new people or ideas.

Have you ever had the urge to throw your computer out of the window? The root cause of such emotional outbursts is often the fact that machines exhibit no empathy. For all the technological advances in computational power, ability to process data on a massive scale and an increasing improvements in AI, the devices and software we interact with on a daily basis remains largely emotionally inert. This could be about to change as a result of dramatic progress in the field of affective computing – the branch of computer science concerned with enabling the recognition, interpretation, processing and simulation of human emotion. The field is not new – it gained prominence through research lead by Rosalind Picard at the MIT media in the 1990s – but recent developments spanning neuroscience, psychology, software development and robotics mean that today's businesses now have the capability to engage employees and customers at an emotional level, creating new opportunities for human machine interaction and symbiotic working.

The arrival of artificial intelligence and its ilk -- cognitive computing, deep machine learning -- has felt like a vague distant future state for so long that it's tempting to think it's still decades away from practicable implementation at the point of care. And while many use cases today are admittedly still the exception rather than the norm, some examples are emerging to make major healthcare providers take note. Regenstrief Institute and Indiana University School of Informatics and Computing, for instance, recently examined open source algorithms and machine learning tools in public health reporting: The tools bested human reviewers in detecting cancer using pathology reports and did so faster than people. Indeed, more and more leading health systems are looking at ways to harness the power of AI, cognitive computing and machine learning. "Our initial application of deep learning convinced me that these methods have great value to healthcare," said Andy Schuetz, a senior data scientist at Sutter Health's Research Development and Dissemination Group.

If we tackle a supervised learning problem, my advice is to start with the simplest hypothesis space first. I.e., try a linear model such as logistic regression. If this doesn't work "well" (i.e., it doesn't meet our expectation or performance criterion that we defined earlier), I would move on to the next experiment. I would say that random forests are probably THE "worry-free" approach - if such a thing exists in ML: There are no real hyperparameters to tune (maybe except for the number of trees; typically, the more trees we have the better). On the contrary, there are a lot of knobs to be turned in SVMs: Choosing the "right" kernel, regularization penalties, the slack variable, ... Both random forests and SVMs are non-parametric models (i.e., the complexity grows as the number of training samples increases).

For a machine learning junkie like me, there was lots to love at Build 2016! In this post, I'll fill you in on the machine learning announcements from Build. In summary, we announced previews of the Microsoft Bot Framework and the Microsoft Cognitive Services (formerly Project Oxford) for adding intelligence to your applications. The Microsoft Bot Framework allows you to build intelligent bots to interact with your users naturally in writing, using text/SMS, Skype, Slack, Office 365 mail, and other popular services. During the keynote, they showed an example Domino's Pizza bot which you could talk to (actually, type to) in natural language to order a pizza.

Scott Aaronson has one of the highest intelligence/pretension ratios I've ever encountered. I wasn't really aware of him before last fall, when I attended a conference at New York University on an ambitious new theory of consciousness, integrated information theory. Most speakers touted IIT or tried to tease out its implications. The striking exception was Aaronson, a boyish (he turns 35 on May 21 but looks younger) computer scientist at MIT (soon leaving for the University of Texas--too bad, MIT!). Although at first he seemed nervous, even jittery, he proceeded to demolish IIT. He focused on a key IIT variable, phi, which denotes the inter-connectivity, or synergy, of the parts of a system. The more phi a system has, the more consciousness it has, supposedly. Aaronson argued--or showed, actually--that IIT's mathematical definition of phi implies that a simple information-storage device, like a compact disc, can be more conscious than a human being. Browsing Aaronson's blog, "Shtetl-Optimized," I discovered that he writes not only about quantum computation, his specialty, but also about artificial intelligence, mathematics, cosmology, particle physics, philosophy… Aaronson has things to say about almost everything. Even when he is at his most technical, he expresses himself in a down-to-earth, funny, self-deprecating and above all clear way. He exudes the spunky enthusiasm and curiosity of a 10-year-old kid, a kid who happens to have a firm grasp of mathematics and physics. He thinks I'm wrong about the end of science, and that's fine with me. Hell, he might be right! I won't say more about him here, because I don't want to embarrass him--or myself--more than I already have, and because he reveals so much of himself in what follows. Warning: this is an extra-long Q&A, but if you read it, I predict, you too will become an Aaronson fan. Come on, that's too high a bar! When I was a kid, I wanted to be the founder and ruler of a rationalist space colony, who also wrote video games and invented the first human-level AI and led a children's liberation movement and discovered the mathematical laws underlying society. On the other hand, as far as childhood dreams go, I have no right to complain. I have a wonderful wife and three-year-old daughter. I get paid to work on engrossing math problems and mentor students and write about topics that interest me, to do all the things I'd want to do even if I weren't getting paid. It's one of those things, like a joke, that dies a little when you have to explain it--but when I started my blog in 2005, it was about my limitations as a human being, and my struggle to carve out a niche in the world despite those limitations. It also gestured toward the irony of someone whose sensibility and humor and points of reference are as ancient as mine are--I mean, I already felt like a senile, crotchety old man when I was 16--but who also studies a kind of computer that's so modern it doesn't even exist yet.

For those that wonder what Google is actually aiming for, Larry Page offers clear direction. He said recently, "Artificial intelligence would be the ultimate version of Google. The ultimate search engine that would understand everything on the Web. It would understand exactly what you wanted, and it would give you the right thing." The key components of the quote are that this AI-based uber-Google would not only know "everything" on the web (basically the sum of all human knowledge) but would also have an insight into the specific needs of each user, presumably even before the need was stated (hence the use of the term "what you wanted"). These two differences are central to the AI leap – essentially aimed at helping Google make the move from a highly sophisticated database search engine that compares text to a system that has a level of "understanding" of the context.

SEOUL--Humanity has fallen to artificial intelligence in checkers, chess, and, last month, Go, the complex ancient Chinese board game. But some of the world's biggest nerds are confident that machines will meet their Waterloo on the pixelated battlefields of the computer strategy game StarCraft. A key reason: Unlike machines, humans are good at lying. StarCraft, created in 1998, is one of the world's most popular computer game franchises. It pits three races against one another: the humanlike Terrans, the slimy insectoid Zerg and a mystical race with psionic powers called the Protoss.

Autonomous cars use a variety of technologies like radar, lidar, odometry and computer vision to detect objects and people on the road, prompting it to adjust its trajectory accordingly. But these tools can drive up the cost of driverless cars, and still aren't as effective as the human brain in visually distinguishing some objects from pedestrians. To tackle this problem, electrical engineers from University of California, San Diego used powerful machine learning techniques in a recent experiment that incorporated so-called deep learning algorithms in a pedestrian-detection system that performs in near real-time, using visual data only. "We're aiming to build computer vision systems that will help computers better understand the world around them," said Nuno Vasconcelos, an electrical engineering professor at the University of California San Diego who led the study, quoted in a story posted by UC San Diego's Jacobs School of Engineering. The findings, which were presented at the International Conference on Computer Vision in Santiago, Chile, are an improvement over current methods of pedestrian detection, which uses something called cascade detection.