It’s shaping up to be a busy summer at Numenta, and I’m pleased to share several news items with you this month.
Highlights from Numenta at Machine Learning and Applied AI events
Our VP Research Subutai Ahmad delivered a talk and poster at a workshop on Uncertainty & Robustness in Deep Learning at ICML (International Conference on Machine Learning) 2019 in Long Beach, California. The talk, which was based on a recent paper titled, “How Can We Be So Dense? The Benefits of Using Highly Sparse Representations,” focused on how adding sparse representations to machine learning systems can increase power efficiency and robustness to noise. If you’re interested in learning more, you can access a recording of Subutai’s ICML talk, the slides from his presentation and the poster he presented. Additionally, Numenta Researcher Lucas Souza joined Subutai at ICML and captured the highlights of the conference in a blog post.
Shortly after ICML, our Community Manager Matt Taylor attended the Re-Work Summit in San Francisco, California. The event comprised three tracks: AI for Good, Applied AI, and Deep Reinforcement Learning. You can see Matt’s video recap of the event here.
Live-streaming our Research Meetings on Twitch
Research meetings have always been the cornerstone of our approach to brain-based machine intelligence. And while we’re constantly striving to make our work open and accessible to as many people as possible, unless you happened to be visiting our headquarters in Redwood City, you couldn’t attend a meeting. Now you can have a front row seat, thanks to Matt’s exploration of Twitch, which he wrote about in a recent blog post. Matt has started live-streaming our research meetings, and we’re excited to be able to engage a broader community. Even if you can’t attend the livestream, we make the recordings available on our YouTube channel, so you can access previous meetings. To join in live, follow Matt’s Twitch channel, where he posts a weekly agenda of upcoming research topics. To watch previously streamed meetings, visit the playlist on our YouTube channel.
New Research Topic on Frontiers in Computational Neuroscience
Recently, Frontiers in Computational Neuroscience announced a new research topic titled How Can Neuroscience Contribute to the Development of General Artificial Intelligence?, and Subutai was named as one of the topic’s editors. The brain has long been a source of inspiration for AI, but it is not always easy for AI researchers to apply its principles. Neuroscience papers often look at very narrow aspects of the brain and are difficult for non-neuroscientists to analyze. The goal of this new research topic is to make core neuroscience ideas accessible to both AI researchers and neuroscientists by generating a number of manuscripts for those interested in biologically-constrained general AI. You can read more about the research topic and submit an abstract or manuscript here.
New Journal Club Meetup Group: Brains@Bay
Are you interested in brain inspired machine learning algorithms? Would you like to discuss the topic with other similarly interested people? One of our researchers has started a new Meetup group called Brains@Bay that’s all about fostering the study and development of machine learning algorithms heavily inspired by neuroscience research. The goal is to meet the 2nd Wednesday of each month to eat, network, discuss papers and view presentations. The first meeting, which will focus on Continuous Learning, will take place at Numenta on July 10 at 7pm. Future locations may vary. To join the group, reserve your spot, and see upcoming themes, visit the Brains@Bay meetup page.
ZDNet interview with Jeff Hawkins on the Thousand Brains Theory of Intelligence
Our Co-Founder Jeff Hawkins was recently interviewed for the Tonya Hall Showon ZDNet and Tech Republic. In the ten minute video, Jeff chats with Tonya about Numenta’s scientific and machine intelligence missions, the Thousand Brains Theory of Intelligence, and how its core principles can be applied to today’s artificial intelligence and machine learning systems to build truly intelligent machines.
Ask Me Anything (AMA)
On May 15 we had a Reddit Ask Me Anything with our Co-founder Jeff Hawkins, VP of Research Subutai Ahmad, and Open Source Community Manager Matt Taylor. If you missed it, you can view the link here. We received some great questions ranging from how the Thousand Brains Theory of Intelligence works to how our research differs from other approaches. The link will stay live on the AskScience subreddit, so if you have additional questions, feel free to post them, and we’ll try to answer as many as possible.
Our commercial license partner Intelletic Trading Systems (ITS) began 100% autonomous, non-human, 24/6 live trading of futures markets on April 7th. The live-trading metrics to date correspond with ITS’s 3-year back-tests and 9-month live simulated-dollar trading. A moderate length live-trading history will enable ITS to proceed with the next phase of their growth, which is to establish an ITS investment fund or alternately share ITS trading signals worldwide directly to client owned trading accounts.
ITS believes that their proprietary cortical learning models based on Numenta’s core algorithms continue to be the best approach to predicting financial time series. Research is underway on constructing ‘first-in-the-world’ multi-level and multi-mode cortical models to further improve the platform’s dynamic response to market, entry and exit conditions. ITS continues to welcome inquiries regarding Intermediate and Series A equity investment from qualified investors. They can be contacted at www.intelletic.com.
Thank you for continuing to follow Numenta.