For 1 or 2 years I searched for a linux distro for AI. It seemed to me that such a distro should exist, because artificial intelligence is a hot topic, and probably many people, like me, would love to dive in and tinker on a ready-made platform. After all, there are many Linux distros tailored to niche interests, why not AI? But I didn't find anything like that. Having once before created a linux distro I decided that maybe I would make another.

Facebook today release new statistics on the growth of chatbots on its Messenger messaging app and also announced new features related to chatbots. Since the chatbot platform launched in April, more than 11,000 bots have been added to Messenger, Facebook vice president of messaging products David Marcus wrote in a Facebook post. For some context, in May, Marcus said that "tens of thousands" of developers were working on Messenger bots. Also, more than 23,000 people have created accounts for the Bot Engine tool from wit.ai, Facebook's natural language processing service for developers, Marcus wrote. And Facebook now lets users rate Messenger chatbots, from one to five stars, and write reviews.

It's a cliché to proclaim that technology is having a disruptive influence on our lives, changing our society and the way people work. And, are employees and organisations prepared for the future of work? A recent event – "The Robots are coming: The future of work"- at the South Bank Centre brought together a panel of experts on robotics and artificial intelligence to debate what awaits our working lives in the near future. I was there to explore the topic and report on the talk. Sabine Hauert (pictured right) is a lecturer in robotics and member of Bristol Robotics Laboratory, an academic centre for multi-disciplinary robotics research in the UK.

"What do I type?" is the big question making chatbots hard to use. So today Facebook Messenger is giving chatbot developers new "Quick Reply" buttons and persistent menu options to make their bots easier to navigate. Messenger bots can also now send videos, audio, GIFs, and files so they can encompass wider range of use cases. People can now rate bots with one to five stars to teach developers how to improve, though there's no word on the chatbot analytics Facebook has promised. And if customers opt in, developers will be able to connect these customers' accounts to Messenger accounts to allow more seamless communication with them.

As government services move online, chatbots may be able to help answer citizen questions. So far, actual government examples are still rare, but speakers at a recent event stressed that the potential use cases are real and widespread. At the DigitalGov University event, "Automatic for the People: AI, Machine Learning and Chatbots for Digital Customer Service in Government" at the General Services Administration headquarters June 28, panelists discussed how advances in artificial intelligence and machine learning, including bots, can expand and improve digital customer service. As more people are using government services online -- and asking questions about those services via digital channels -- the number of daily online queries to some agencies have already hit hundreds per day. "What happens when it becomes thousands of questions a day?" Justin Herman, the lead for open government at GSA's Technology Transformation Service, asked.

NEWS ANALYSIS: The use of machine learning to identify suspicious online activity is a new and important capability in securing the network, but privileged users were the weak point until now.

Google's Chief Internet Evangelist Vinton Cerf, left, and Joachim Müller-Jung of the FAZ newspaper's science and nature section, right. But Vinton (Vint) Cerf, a Google vice president and Chief Internet Evangelist, was also on the panel. Cerf had fun regaling the audience with stories of Google's self-driving experiments and how one of their cars had hit a bus but that it was "only at about 3 kilometers per hour (2 mph)," and how in other cases Google's autonomous cars had been rear-ended by other cars, driven by humans. To suggest code is neutral and without philosophy or ethics is to suggest a future where we can happily say to bereaved parents, "I'm terribly sorry that my car ran over your child, but it was the car's fault, I was in the back having sex."

It was an accident waiting to happen. Up until then, I had been rather bored. So it was Wednesday, and the third "Press Talk" at the 66th Lindau Nobel Laureate Meeting. And the topic was artificial intelligence (A.I.). Müller-Jung, who's the head of science and nature at the German daily newspaper "Frankfurter Allgemeine Zeitung," had repeatedly said during his long and winding introduction, "We're not going to talk about self-driving cars or'rogue A.I.' here!"

Tesla announced today that the National Highway Traffic Safety Administration has opened an investigation into a recent fatal crash of a Model S with the company's Autopilot feature activated. The accident took place on May 7th in a small West Florida town called Williston. The Florida Highway Patrol is also conducting its own investigation of the accident, according to a public affairs officer there. The same officer reported that Tesla has, since the fatal accident in May, sent engineers down to Ocala, Florida to assist investigators in accessing data they needed to evaluate the causes of the crash. Tesla offered an account of the event in a blog post titled "A Tragic Loss" that went up today, detailing the crash, an "extremely rare circumstance," which occurred on a divided highway.

Our second model, called very deep convolutional neural net (or CNN), has its origins in image classification [4]. Speech can be viewed as an image if we consider the spectral representation of the audio signal with the two dimensions being time and frequency. As opposed to the classic CNN architectures employed in our previous system [5] that have only one or two convolutional layers with large (typically 9-by-9) kernels, our very deep CNN [6] has up to ten convolutional layers with small 3-by-3 kernels which preserve the dimensionality of the input. By stacking many of these convolutional layers with Rectified Linear Units nonlinearities before pooling layers, the same receptive field is created with less parameters and more nonlinearity. These two models which differ radically in architecture and input representation show good complementarity and their combination leads to additional gains over the best individual model.