Summary: Despite our concerns about China taking the lead in AI, our own government efforts mostly through DARPA continue powerful leadership and funding to maintain our lead. Here’s their plan to maintain that lead over the next decade.
Think all those great ideas that have powered AI/ML for the last 10 years came from Silicon Valley and a few universities? Think again. Hard as it may be to admit it’s the seed money in the billions that our government has spent that got pretty much all of these breakthroughs to the doorway of commercial acceptability.
Dozens of articles bemoan the huge investments that China is making in AI with the threat that they will pull ahead. That ignores the substantial investments our own government is making to stay in the lead.
It’s not popular to give Washington much credit for intellectual leadership so let’s be specific; it’s mostly DARPA, the Defense Advanced Research Projects Agency. And while DARPA may lead and set the agenda, few of us think about the work being done at the national laboratories like Lawrence Berkeley of which there are 10 plus an additional 7 between DOE and NNSA.
What most Americans are unaware of is that in addition to these 17 there are an additional 25 Federally Funded Research and Development Centers (FFRDCs) that gives our government-directed research goals some really hefty muscle.
It’s not about commercialization that DARPA spends its money (though in many cases it’s about national security). It’s about setting the medium range goals that become the seed capital that underlies most of our current AI/ML.
Autonomous vehicles came out of the DARPA challenges a decade ago. Neuromorphic chips got their start there also in about 2006. Their voice powered assistant SIRI was acquired by Apple in 2011. Look closely and you’ll find the roots of pretty much all our AI/ML watered by some early DARPA care and feeding.
Investment for the Next 10 Years
Forward goals and funding follows the fed’s AI strategy called the ‘American AI Initiative’ signed by President Trump in 2018. Most of the new $2 billion multi-year funding will be through DARPA and has at least two different roadmaps forward. DARPA’s Artificial Intelligence Exploration (AIE) program seeks to expand fundamentals and their ‘AI Next’ program which appears to be more focused on breakthrough applications.
There are five major areas of investigation described by DARPA.
New Capabilities: Includes more than 60 existing programs covering such areas as real-time analysis of cyber-attacks, detection of fraudulent imagery, construction of dynamic kill-chains for all-domain warfare, human language technologies, multi-modality automatic target recognition, biomedical advances, and control of prosthetic limbs. This area also covers more business like applications such as enabling automation of DODs business processes.
Robust AI: Seeking to overcome our lack of understanding about the failure modes of AI to ensure reliable performance.
Adversarial AI: Overcoming the vulnerability that many ML systems can be duped by changes to inputs that would never fool a human. Also detecting and preventing intentional corruption of training data and the direct threat of cyber-attack.
High Performance AI: Hardware and software solutions to dramatically reduce power consumption and increase learning efficiency allowing deployment of low-power and low-latency applications to edge devices. Eliminating vulnerabilities related to streaming data and dramatically reducing requirements for labeled training data.
Next Generation AI: Enabling AI applications to explain their actions, acquire and reason with common sense knowledge, and incorporate causality in their logic.
DARPA has both a philosophy and reputation for disruption. In addition to more traditional projects they will fund “a series of high-risk, high payoff projects where researchers will work to establish the feasibility of new AI concepts within 18 months of award.”
“We are harvesting the intellectual fruit that was planted decades ago. That’s why we’re looking at far forward challenges – challenges that might not come to fruition for a decade” says John Everett, deputy director of DARPA’s Information Innovation Office.
About the author: Bill is Contributing Editor for Data Science Central. Bill is also President & Chief Data Scientist at Data-Magnum and has practiced as a data scientist since 2001. His articles have been read more than 1.5 million times.
He can be reached at: