What is Machine Discover & Invention?
Cognitive Computing methodology is rapidly growing, is becoming increasingly powerful, and is being widely embraced by major corporations, such as IBM, for many enterprise applications. More recently these approaches are being applied in a growing number of “hard” science and technology R&D areas. ETC is advancing Cognitive Computing methods to not only accelerate and enhance traditional science and technology research and development productivity, but to provide innovative new approaches to accelerate the pace of discovery - this use of powerful new cognitive computing capability we have named Machine Discovery and Invention (MD&I). Some view the approaches and processes used in MD&I as Artificial Intelligence (AI), a field that has seen much development over the past fifty years. In order to increase clarity, our approach doesn’t attempt to understand or define intelligence; instead, it looks to outcomes, measured by actually finding new knowledge that can be objectively described as discoveries or inventions. The term ‘Machine’ is meant here to refer to algorithmic, automated, repeatable approaches, typically characterized as being performed by a computer system. The computers used for MD&I may be composed of one or many processors, and can range from simple lap-tops to general-purpose CPUs, to application-specific graphic/vector manipulation chips, to AI accelerators, and eventually perhaps to neuromorphic processors and quantum computers.
In general, our goal is to use Machine Discovery & Invention to enable two things; first, to enhance and accelerate discoveries and inventions in areas that have seen a dwindling of significant new discoveries and inventions; and second, discovery and inventiveness in totally new areas.
Another way of visualizing Machine Discovery and Invention is to focus and accelerate the high-level functions of Goal-Oriented R&D and Open-Ended Exploration, while combining the benefits of both. Goal-Oriented R&D generally starts with some specific purpose in mind, while Open-Ended Exploration is less prescriptive. You can start out along one path and stumble into something interesting that really belongs along the other path. The key to many important new discoveries however is to notice and appreciate unusual and unexpected phenomena or patterns or roadblocks, as those may be strong hints of some new discovery or innovation approach. In MD&I, this ability for “noticing and appreciation” no longer relies only on human expertise; it will be built into the autonomous methodology.
OUR WORK IN MACHINE DISCOVERY & INVENTION
Here are some of the work we have done so far, what we are currently working on, and what we plan to do moving forward.
In FY15, under the Office of Naval Research (ONR) sponsorship, ETC completed an initial exploratory study on the potential of Literature Based Discovery (LBD) to address medical issues related to undersea diving. The study and recommendations were documented in the AUTOMATED BIOMEDICAL KNOWLEDGE BASED DISCOVERY FINAL TECHNICAL REPORT.
In FY19, ETC completed an MD&I research project under a grant from the Office of Naval Research. In this work, we developed an approach to accelerate the discovery and characterization of new energetic molecules. In this work, we developed computational methods to autonomously develop new candidate energetic molecules, and to use machine learning to very rapidly predict what their energetic performance would be. Results from this project were reported in five journal publications and were documented in “Final Report for ‘Use of Machine Discovery & Invention (MD&I)
to Accelerate Undersea Weapons and Energetics Technology Development”.
In FY20, again with Office of Naval Research sponsorship, we began a new 3-year MD&I project to develop and apply Machine Learning and Natural Language Processing methods to autonomously characterize and help invent new explosive formulation fills, intended to improve the performance of undersea warheads.
In the longer term, we look to apply the experience gained from our prior efforts and the Energetic Materials MD&I effort to larger and broader scale initiatives. A key thrust will be to formulate a national level MD&I initiative of a general nature.