When I wrote up'Asynchronous methods for deep learning' last month, I made a throwaway remark that after Go the next challenge for deep learning systems would be to win an esports competition against the best human teams. Can you imagine the theatre! Since those are team competitions, it would need to be a team of collaborating software agents playing against human teams. Which would make for some very cool AI technology. Today's paper isn't quite at that level yet, but it does show that progress is already being made on playing first-person shooter (FPS) games in 3D environments.

A standard 6-person meeting room (C) is adjacent to the brainstorm room covered with whiteboards (D). A variety of meeting spaces is an essential part of the WeWork experience, but finding the right combination can be challenging. How many meeting rooms do you need in an office? It's a simple question, but one that is very difficult to answer. Even experienced architects and designers struggle to allocate the correct number of meeting spaces, relying mainly on rules of thumb and intuition to overcome the lack of empirically verified guidance.

Online learning is becoming the norm in universities across the globe, bringing sweeping changes to the way we learn. But earlier this year on online graduate class at Georgia Tech took things a stage further. "Our Teaching Assistants are getting bogged down answering routine questions," said Ashok Goel, who teaches a graduate science course. Students in the class typically post 10,000 messages a semester on the Piazza forum for the course, many of which are either variations on a theme or simple logistical questions. To address this problem, Ashok turned to IBM Watson to create a virtual TA called Jill Watson who was trained on 40,000 posts and released to the wild on the live forum in March as an addition to the other eight TAs.

Seeing the long queues outside ATMs and the confusion and chaos about the availability of cash in the wake of the demonetisation, two engineering graduates from Coimbatore's Government College of Technology decided to do something to help the struggling people. "With the ongoing mess in the country, we thought'why not build something to help ease the situation'. Though there are other interfaces that dispense similar information, ATMBot is different because it is crowd-funded. Users need not download an extra application on their phone to use it," says Abishek Muthian, who founded the city-based start-up Timebender Technologies India Private Limited along with Aravindhan Ramachandran. After his graduation, Abishek decided to venture out on his own.

At Inbound 2016, HubSpot's co-founders Brian Halligan and Dharmesh Shah entertained 19,000 attendees with their take on the past and future of marketing. Here's what I learned from their keynote presentation and a brief interview. So predicts Halligan, adding "in five years, you will do a lot less navigating through apps and more just asking questions and chatting back and forth with bots… the next thing you know, we like it and it's easier and more efficient than waiting for the sales rep to call you back." Shah notes that businesses started building websites in the 1990s so they can answer customer questions 24/7. "Soon," he says, "they will start building bots. They won't replace the websites, but they will power them. The shortest time between a customer question and the answer will be a bot. It's not human vs. bot, it's human to the bot powered."

On Nov 7th, the AI World Expo kicked off in a week that had many international conferences occurring. After making the choice to forego Web Summit in Lisbon for this event, it was evident that the level of quality of content, attendees and hands on innovation in the #AIWorld Expo made it the ideal event for AI innovators, thought leaders and adopters. The concentration of hand-selected vendors who were pre-screened rigorously, and the agenda made the event a great launchpad to build new relationships, get some face time with pioneers, and spend quality time discussing the interests of each vendor seeking a competitive edge. It was an interesting outcome and dynamic for a group of people fascinated with and specializing in artificial interactions, chat bots, autonomous learning and IoT! To kick things off, highly engaging panelists representing the venture community discussed how difficult it was to tap into the mainstream of AI.

A key ambition of BenevolentAI is to utilize artificial intelligence in revolutionizing the drug development process. Now, with a licensing agreement with Janssen, it is one significant step closer to realizing that dream. With assistance from Johnson & Johnson Innovation's Centre in London, the British AI company has acquired an undisclosed number of novel clinical stage drug candidates, together with their related patents. The company used its AI platform to evaluate the potential of these small molecule compounds and found some could be promising candidates for hard-to-treat diseases. "The compounds come with a wealth of clinical and biological data that enables BenevolentAI to have further insights into the biology of diseases," said BenevolentAI Bio's CEO Jackie Hunter.

Machine learning and data analysis are two areas where open source has become almost the de facto license for innovative new tools. Both the Python and R languages have developed robust ecosystems of open source tools and libraries that help data scientists of any skill level more easily perform analytical work. The distinction between machine learning and data analysis is a bit fluid, but the main idea is that machine learning prioritizes predictive accuracy over model interpretability, while data analysis emphasizes interpretability and statistical inference. Python, being more concerned with predictive accuracy, has developed a positive reputation in machine learning. R, as a language for statistical inference, has made its name in data analysis.

Google may have found a way to use machine learning technology to help millions of deaf and hearing-impaired people better understand what people are saying to them. Researchers from Google Inc.'s DeepMind artificial intelligence project, which built the boardgame-playing AlphaGo that managed to successfully defeat one of the top Go players in the world, have teamed up with peers at the Oxford University to create an AI system that's able to outperform professional lip-readers after training itself on thousands of hours of BBC videos. New Scientist reports that in tests, a human lip-reader who provides services for the U.K. courts was able to correctly decipher only about a quarter of words spoken when shown a random sample from 200 BBC video broadcasts. However, DeepMind's AI system was able to decipher almost half of the words from the same sample videos. In addition, the AI was able to annotate 46 percent of the words without error, compared with just 12 percent by the human lip-reader.

The O'Reilly AI Conference is returning to New York June 26–29, 2017 to explore the most essential and intriguing topics in intelligence engineering and applied AI. The program will cover the latest developments in tools, algorithms, and architectures, applications such as finance and robotics, novel interfaces like bots, plus much more. We're looking for compelling case studies, technical sessions, tear-downs of both successful and failed AI projects, technical and organizational best practices, and more. See online for a list of suggested topics, but feel free to recommend others because we always love to be surprised. See our tips on how to submit a great proposal.