Panel – U.S. Military’s Mad Science Revealed, presented by Christie Nicholson and John Pavlus
Summary:
For more than 50 years the mad scientists at the Defense Advanced Research Projects Agency—aka DARPA, the outrageous research arm of the Pentagon—have been launching the most disruptive technologies on earth, living up to their mantra of “high risk—high payoff.” We have DARPA to thank for the personal computer, the Internet, the Berkeley Unix system, most of NASA, and countless crazy military innovations. Their mission is to think beyond the possible and forever be three decades ahead. In this talk we will dig into, and present the relevant parts of, DARPA’s $3 billion-dollar budget, pulling out the most amazing and most-likely-to-reach-fruition projects. Think electromagnetic bazookas, telepathic soldiers, ape-inspired robots, memory chips in brains, shapeshifting planes and boats. It might sound like sci-fi, but given its inspired history it seems that analyzing DARPA’s current projects will give us one of the clearest views into our future reality. Fasten your seat belts.
Notes:
DARPA – Defense Advanced Research Project Agency
- Bio-Inspired Agile Robots
- DARPA’s attempt to solve problems using robots
- Problem- Robots tend to be clumsy. DARPA is working on agile robots, with robots such as Big Dog – built by Boston Dynamics
- Goal – design a robots that is autonomous, agile, and designed for HUMAN environments
- Petman – walking robot created by Boston Dynamics
- If an robot is going to look and move like an animal, it needs to have similar inteligence and processing powere to allow it to move intelligently, instead of like a Roomba. Gotta be smart enough not to run into a concrete wall.
- Deep Learning
- A technique for learning like a human would learn.
- Supervised learning – give the system a few set examples, and find the specific date. Brittle system
- Unsupervised learning – Next generation learning methodology. Show system a few representations, and it forms an abstraction of the idea of the image, and can be much more flexible and powerful.
- Object Recognition
- MSEE – Mathematics of Sensing, Exploitation & Execution
- Algorithm which captures all the complexity of taking in raw sensory input, applicable to any sensor machine, and make sense of all sensory input.
- Brain Computer Interface
- Andy Schwartz’s Motor Lab, University of Pittsburgh