Get Over the I nsecurity! Ed Lazowska Depart ment of Comput er - - PowerPoint PPT Presentation

Get Over the I nsecurity!

Ed Lazowska Depart ment of Comput er Science & Engineering Universit y of Washingt on

Key point s

Don’t get hung up on t rying t o be a “pure science”

The f act t hat much of what we do is usef ul is good, not bad Sur e, t he physicist s did The Mot her Of All Demos back in 1945, but t hey’re in t he crapper t oday – now, t hey envy us!

We are at t he cent er of everyt hing

There are incredible opport unit ies f or “peer t o peer” int ellect ual advancement

“J ust say no” t o t hose who want somet hing else f rom you – cor por at e or academic But r ecognize t hat ever y par t y in a collabor at ion needs t o “pay some dues”

Beware of having a narrow view of what const it ut es comput er science

Science vs. engineering

Science

Describe, explain

Engineering

Design, build, evaluat e “An engineer can do f or a dime what any f ool can do f or a dollar ”

Much of comput er science is engineering – celebrat e t his!

“Engineering research”: oxymoron?

“Fundament al research” and “applicat ion- mot ivat ed research” are compat ible

Tradit ional view

Fundamental research Applied research

Alt ernat ive view

Concern with fundamentals Concern with use

Edison Pasteur; much of biomedical and engineering research Bohr

Some UW examples in t he bio space

Comput at ional molecular biology LabScape – embedded syst ems t o inst rument biot ech laborat ories Neurally-inspired comput ing

Comput at ional Molecular Biology

Collabor at or s: Lee Hood, Maynar d Olson, Phil Gr een Facult y: Dick Karp, Mart in Tompa, Larry Ruzzo, Rimli Sengupt a Post docs: Amir Ben-Dor, Benno Schwikowski Complet ed Ph.D. st udent s: Br endan Mumey (U

Mont ana), J er emy Buhler (WashU), Ka Yee Yeung (UW Microbiology), Agat ha Liu (I BM), Saur abh

Sinha (Rockaf eller U), Mat hieu Blanchet t e (McGill), Emily Rocke (UW Genome Sciences) Cor por at e int er act ions: Zymogenet ics, I mmunex, Roset t a, I nst it ut e f or Syst ems Biology

Gaet ano Bor r iello Gaet ano Bor r iello Depar t ment of CS&E Depar t ment of CS&E Univer sit y of Washingt on Univer sit y of Washingt on Seat t le SAGE Gr oup Seat t le SAGE Gr oup 14 Sept ember 2000 14 Sept ember 2000

The The Port olano Port olano Expedit ion Expedit ion in I nvisible Comput ing in I nvisible Comput ing

port olano port olano.cs cs.washingt on washingt on.edu edu

P rincipal Themes P rincipal Themes

• I nvisibilit y

I nvisibilit y

not enough t o be mobile, pervasive, ubiquit ous, et c.

not enough t o be mobile, pervasive, ubiquit ous, et c.

user’s at t ent ion is t he valuable resource

user’s at t ent ion is t he valuable resource

minimize user conf igurat ion/ maint enance/ int eract ion

minimize user conf igurat ion/ maint enance/ int eract ion

robust , reliable, saf e, and t rust wort hy

robust , reliable, saf e, and t rust wort hy

devices, middle

devices, middle-

• ware, and “applicat ions”

ware, and “applicat ions” services services

• Act ive f abric

Act ive f abric

plug

plug-

• and

and-

• play, discovery,

play, discovery, composabilit y composabilit y

dat a

dat a-

• cent r ic, het er ogeneous, act ive net wor king

cent r ic, het er ogeneous, act ive net wor king

dat a and code mobilit y

dat a and code mobilit y

self

self -

• or ganizing, self
• r ganizing, self -
• updat ing, self

updat ing, self -

• monit oring syst ems

monit oring syst ems

act ive dat abases and inf ormat ion management

act ive dat abases and inf ormat ion management

• Ext ernal user communit y

Ext ernal user communit y

publicat ion

LabScape LabScape -

• one of our driver applicat ions
• ne of our driver applicat ions
• Biology is a hard science wit h a sof t inf rast ruct ure

Biology is a hard science wit h a sof t inf rast ruct ure

• capt ure and use of knowledge is key

capt ure and use of knowledge is key

• f rom loosely connect ed t o highly int egrat ed collaborat ion

f rom loosely connect ed t o highly int egrat ed collaborat ion

• invisible inf rast ruct ure f or building knowledge base

invisible inf rast ruct ure f or building knowledge base

I nt erpret Experiment Hypot hesize I nt erpret Experiment Hypot hesize Descript ive Model Experiment Manager

knowledge base knowledge base

Event Capt ure in Labscape Event Capt ure in Labscape

Neurally Neurally Inspired Inspired Computation Computation

Chris Diorio

Computer Science & Engineering University of Washington diorio@cs.washington.edu

Nature is telling us something...

Can add numbers together in

nanoseconds

Hopelessly beyond the capabilities of brains

Can understand speech trivially

Far ahead of digital computers …and Moore’s law will end

Problem: How do we build circuits that learn

One approach: Emulate neurobiology Dense arrays of synapses

• utput = ∑ W

X

j j j 2

X1 X2 learn signal input vector X error signal

• utput = ∑ W X

j j j 1

synapse W11 synapse W12

learn signal error signal

synapse synapse W21 W22

Silicon synapses

n– electron tunneling p electron injection n+ n+ n+

Silicon Synapse Transistor Charge Q Sets the Weight

p– substrate 1 2 3 4 5 10-11 10-10 10-9 10-8 10-7 10-6 10-5

control-gate–to–source voltage (V) source current (A)

Q1 Q2 Q4 Q5 Q3

floating gate (charge Q)

Use the silicon physics itself for learning Local, parallel adaptation Nonvolatile memory

Silicon synapses can mimic biology

–10 10 20 30 40 50 1 2 3 4 5

time (min) synapse source currents (nA)

Biological Synapses Silicon Synapses

Mossy-fiber EPSC amplitudes plotted over time, before and after the induction of LTP. Brief tetanic stimulation was applied at the time in-

• dicated. From Barrionuevo et al., J. Neurophysiol. 55:540-550, 1986.

Synapse transistor source currents plotted over time, before and after we applied a tetanic stimulation of 2×10 5 coincident (row & column) pulses, each of 10 µs duration, at the time indicated.

Local, autonomous learning

Synaptic circuits can learn complex functions

Synapse-based circuit operates

• n probability distributions

Competitive learning Nonvolatile memory 11 transistors 0.35µm CMOS Silicon physics learns “naturally”

Silicon learning circuit versus software neural network

Both unmix a mixture of Gaussians Silicon circuit consumes nanowatts Scaleable to many inputs and dimensions

true means circuit output software neural network

value (V) number of training examples 2000 4000 0.2 0.4 0.6 0.8 1 1000 3000

Technology spinoff: Adaptive filters

Synapse transistors for signal processing

~100× lower power and ~10× smaller size than digital Mixed-signal FIR filter

16-tap, 7-bits 225MHz, 2.5mW Built and tested in 0.35µm CMOS Adjust synaptic tap weights off-line

FIR filter with on-chip learning

64 taps, 10 bits, 200MHz, 25mW In fabrication in 0.35µm CMOS On-line synapse-based LMS

Startup company: Impinj

Chris Diorio (UW) and Carver Mead (Caltech) Self-tuning analog computing implemented in standard digital CMOS

processes (e.g., TSMC) for telecommunications applications (filtering, DSP, etc.)

Potentially a factor of 500 power savings, plus the ability to fully

integrate analog and digital on the same die

Problem: How to study neural basis of behavior

Measure neural signaling in intact animals

Implant a microcontroller in Tritonia brain

Tritonia is a model organism

Well studied neurophysiology 500µm neurons; tolerant immune response Work-in-progress

• B. Brain with implanted chip: Dorsal view
• A. Tritonia and seapen

Images courtesy James Beck & Russell Wyeth

MEMS probe tip, amplifier brain battery tether memory microcontroller, A/D, cache visceral cavity Tritonia diomedea

An in-flight data recorder for insects

An autonomous microcontroller “in-the-loop” Study neural basis of flight control

Manduca Sexta or “hawk moth”

Participants

Chris Diorio and students from CSE Karl Bohringer and students from EE (MEMS probes) Tom Daniel and students from Zoology Dennis Willows and students from Friday Harbor Labs Funding from Packard, DoD MURI, NSF, DARPA, many

• thers

Key point s

Don’t get hung up on t rying t o be a “pure science” We are at t he cent er of everyt hing There are incredible opport unit ies f or “peer t o peer” int ellect ual advancement Beware of having a narrow view of what const it ut es comput er science