Ron Ben Tsvi
DARPA Funding Opportunity: Virtual Intelligence Processing (VIP)
The U.S. Federal Government is by far America's biggest investor in R&D projects in both the private and public sectors. And it's not even close. DARPA alone invests $3.4B+ ANNUALLY without taking back any equity or making a claim to the IP.
DARPA is issuing an Artificial Intelligence Exploration (AIE) Opportunity inviting submissions of innovative basic or applied research concepts in the technical domain of third wave AI exploring radically new and heretofore under-explored, brain-inspired and massively-scalable computing methodologies for novel machine learning approaches that have the potential to support solutions to real-world problems impacting the Department of Defense.
Project Details: Phase 1: up to $675,000 (12 months) Phase 2: up to $325,000 (6 months) Due Date: July 15, 2019
VIP is an 18 months program divided into a 12-month Phase 1 and a 6-month Phase 2.
Phase 1 performers will develop novel algorithms based on unexplored mathematical models and validate the potential for 10x improvement in energy efficiency and data rate handling capability for a performer-defined DoD relevant challenge problem.
Phase 2 performers will demonstrate a proof-of-concept for a “real world” application of the algorithms developed in Phase 1 by applying them to real-world data sets and will propose a possible hardware architecture for the chosen challenge problem. Phase 2 will also explore scaling projections of performance guiding future hardware developments.
Our best model for intelligence is ourselves, and thus next generation AI will benefit from computing models and processing architectures that can support robust and massive parallelism such as is the case in the human neocortex.
In order to reverse engineer the human brain we need to apply new mathematical models for computing that are complete and transparent and can inform next generation processors that are better suited for third wave AI.
Proposers should address the fundamental performance limits of novel algorithms based on unexplored mathematical approaches when applied to a relevant DoD application. Specifically, proposers must describe their novel approach in terms of accuracy, robustness to environment, data and computational requirements as well as the challenges that need to be overcome for its implementation. They must also quantitatively compare the chosen approach with current state of-the-art for the same application. The program is also interested in hardware simulations but not in actual hardware implementation.