As 2018 comes to a close, I’m pleased to share several exciting updates with you.
Exploring the Impact of the Thousand Brains Theory on Machine Intelligence
I recently continued my yearly tradition of sitting down with Jeff to reflect on the past twelve months and discuss Numenta’s goals and priorities for the coming year. This year, Subutai joined to share some exciting developments for 2019. Longtime followers of Numenta will know that we have always had a dual mission: to reverse engineer the brain and apply that knowledge to the creation of machine intelligence. For the past year and a half, we have focused almost exclusively on the scientific mission. As Jeff wrote about in this blog post last July, we shifted our resources into research to accelerate our progress on brain theory. The result of that shift led to the development of our framework for intelligence and cortical computation, the Thousand Brains Theory of Intelligence, which we formally announced in October. The Thousand Brains Theory proposes that rather than learning one model of an object or concept, the brain builds many models of the same thing. Each model is built in a different way – with different types of inputs from different senses. The models vote together to reach a consensus on what they are sensing, and the consensus vote is what we perceive. It’s as if your brain is thousands of brains working simultaneously!
As we look forward to 2019, we are rebalancing our efforts between our twin missions. On the theoretical side, Jeff will continue to focus on the neuroscience research. On the pragmatic side, Subutai will start a new effort to explore how the Thousand Brains Theory could have a positive impact in the field of machine intelligence. The timing of our return to the machine intelligence mission aligns nicely with the recent realizations that have come out of the deep learning world. Though there are many impressive practical applications, some leaders in the field recognize the fundamental limits to the current approach and have come to the realization that neural nets will not be the path to true machine intelligence. We believe the Thousand Brains Theory can help address many of those limitations. We don’t know where our exploration will lead, but we’re encouraged about the possibilities. We plan to share news of our progress in the coming year.
HBS Alumni feature on Donna Dubinsky and Numenta
Our CEO, Donna Dubinsky, was profiled in this month’s edition of the Harvard Business School Alumni Bulletin, in a piece titled, "Source Code.” This in-depth feature looks at Donna’s career of ushering in different waves of tech revolution, and how that led her to the challenge of “reinventing artificial intelligence—by reverse engineering the human brain” at Numenta. The piece also does a nice job of painting a day in the life at Numenta, the approach we take to neuroscience research, and the ways in which we believe our work will impact AI. As Donna points out in the article, “Current AI is based on 50-year-old algorithms that have been supercharged with more data and computing power. It’s not that they aren’t accomplishing interesting things—they are. But they’re not doing it in a way that your brain does it. So the next big thing is going to be, how can we make computing more brain-like?”
Cortical.io has been focused on their enterprise solution targeting contract intelligence, which incorporates a neuroscience-based algorithm that understands the meaning of text rather than keywords. You can read more about why the time is ripe for contract intelligence in their latest post.
Grok has been busy growing and innovating in the last couple of months since launching Grok 3.0 this summer. In addition to using Numenta’s HTM algorithm for detecting anomalies in time series data, Grok now has the ability to ingest and learn from system events and logs. Additionally, a new microservices-based backbone, capable of scaling to accommodate very large and diverse learning and prediction paradigms, has allowed for new and interesting case studies, including the use of supervised classification to accurately predict emerging IT Incidents. Learn more about incident detection with Grok at https://www.grokstream.com/.
Our commercial licensee Intelletic Trading Systems (ITS) also has had a successful year. We first reported on their use of HTM to predict short term movement in financial markets in May, and then followed up with a blog interview with ITS’s CEO Bill Zemlak in November. ITS continues to have success with simulated real-time trading, where five months of results to date confirm the metrics and returns achieved in their three-year backtest. ITS plans to begin real-time trading of S&P futures on a funded account in early 2019.
From all of us at Numenta, we wish you Happy Holidays and a safe and healthy New Year. Thank you for continuing to follow Numenta.