New Directions for DARPA

1
New Directions for DARPA
Beyond Strategic Computing

• ISAT Ad Hoc Working Group on DARPA Futures
• Initial Draft July 12 2000
• Update July 15 2000 x 2

2
Beyond Strategic Computing

• For 18 years, DARPA has successfully pursued the
  Strategic Computing vision
• Building on this foundation, we are prepared to
  confront the critical challenges facing us today.
  These include
• Augmented cognition
• High-confidence systems and software
• New computational substrates
• Computing and biology

3
The DARPA Strategic Computing Program (1983)
develop a broad base of machine intelligence
technology to increase our national security and
economic strength
Major Goals
autonomous systems pilots associate, battle
management
Military Applications
natural language, vision, speech, expert
systems, navigation, planning and reasoning
Intelligent Functional Capabilities
high-speed signal processing, general purpose
systems, symbolic processors, multi-processor
programming and operating systems
HW/SW System Architecture
silicon and GaAs technology VLSI systems
Microelectronics
networks, research machines, rapid machine
prototyping,implementation systems foundries,
interoperability protocols, design tools
Infrastructure
New-Generation Technology A Strategic Plan for
its Development and Application to Critical
Problems in Defense, DARPA, 1983.
4
Augmented Cognition

• Challenge The volume of information and overall
  complexity of warfighting continue to grow at a
  rapid pace, in stark contrast to human cognitive
  abilities, which remain static
• Memory
• Attention
• Sensory bandwidth
• Comprehension
• Visualization abilities
• On the other hand, computational capabilities
  have continued to grow rapidly

Apply computational power to support / augment
cognitive skills, bolstering limited cognitive
resources
5
Augmented Cognition

• Opportunity Designs and methods that leverage
  new understanding about characteristic
  limitations in cognition
• Enhance memory
• Support the analyst with data analysis,
  discovery, visualization
• Automate new aspects of problem solving and
  filtering
• Modulate, triage communications, information
• Develop new visualization techniques to enhance
  understanding, increase effective human-computer
  bandwidth
• Extend abilities to monitor, control
  semi-autonomous systems, robots

6
High-Confidence Systems and Software

• Challenge Todays systems are fragile, difficult
  to compose and maintain
• Non-robust
• Non-adaptive
• Untrustworthy

• Point failures bring down systems
• Difficult, costly to compose useful systems from
  multiple components
• Poor or nonexistent means for building reliable
  systems from necessarily unreliable components
• Poor understanding of vulnerabilities,
  performance under characterized and
  uncharacterized attacks
• No clear history, pedigree on data, code

7
High-Confidence Systems and Software
Opportunity Develop hardware, software,
algorithms, and overall architectures with a more
fundamental approach to security and fault
tolerance

• Methods for integrating unreliable components to
  create reliable systems as foundation of robust
  computing
• Methods for addressing, leveraging, harnessing
  distributed components and resources.
• Redundancy via replication and efficient restart
  machinery
• Develop new approaches to model, understand,
  control, react to emergent behaviors in complex
  computational systems at baseline and when
  stressed
• Pursue understanding of robustness and adaptation
  exhibited by biological systems for relevance to
  computing

Consider architectural issues spanning multiple
layers of networking, software, computation,
hardware
8
New Computational Substrates

• Challenge We are nearing the end of exponential
  growth in processor performance using existing
  technologies
• The current silicon / fabrication paradigm leaves
  us within sight of a flatlining of the Moores
  law curve
• Atomic dimensions will limit scaling of silicon
  technology

9
New Computational Substrates

• Opportunity Disruptive technologies for
  computing, based upon new computational
  substrates, are potentially on the horizon
• Substrates
• Computational fabric
• Smart matter, MEMS
• Quantum
• Biological approaches to computing
• DNA
• Molecular electronics
• Microbial robotics
• Engineered genetic networks
• Combinations of the above

10
Computing and Biology

• Challenge An explosion of knowledge about
  biological systems suggests that multiple
  disruptive technologies lay waiting, yet we
  remain mired in large amounts of data and limited
  computational tools and models
• Biological systems represent powerful
  architectures for sensing, processing, actuating,
  and fabrication, and for managing complexity in
  an elegant manner
• IT, computational analysis is essential for
  cracking key challenges in biology
• Increased understanding and applications of
  knowledge about biology will likely have
  significant defense implications
• Increased understanding of biology will likely
  reveal new approaches to computing and
  materialswith significant defense implications

11
Computing and Biology

• Opportunity Apply computation, modeling to
  pursue conceptual foundations of biological
  approaches to
• robustness, uncertainty, adaptation, inference,
  replication, repair, communication,
  transformation, distribution, self-assembly,
  fabrication, ontogeny / development, identify
  friend vs. foe
• Key conceptual challenges include understanding
• Modeling of genetic circuits
• Proteomics
• Embryogenesis / development
• Evolvable systems
• Engineered biology

12
Beyond Strategic Computing
develop a broad base of machine intelligence
technology to increase our national security and
economic strength
Major Goals
2020 Application Vision TBD via an ISAT study
Augmented cognition
High confidence systems and software
HW/SW System Architecture
Computing, Microsystems, and Biology
New Computing Substrates from Electronics,
Microsystems, CS