1. Can We Secure the Internet? Prof. Eric A. Brewer UC Berkeley Berkeley in Silicon Valley March 1, 2003

2. Can we reduce the motivation for attack? Prof. Eric A. Brewer UC Berkeley Berkeley in Silicon Valley March 1, 2003

3. Adlai Stevenson, July 1965 We travel together, passengers on a little space ship … all committed for our safety to its security and peace; We cannot maintain it half fortunate, half miserable, half confident, half despairing, half slave … half free in a liberation of resources undreamed of until this day. No craft, no crew can travel with such vast contradictions. On their resolution depends the survival of us all.

4. Hypothesis 1  Current IT projects “trickle down” first- world technology:  Too expensive  Assumes infrastructure, power  Assumes IT knowledge and support  Assumes literacy  We can directly attack these issues

5. Hypothesis 2  Thousands of IT projects, but  Focus on devices not infrastructure  No single project can afford to build infrastructure, but all of them would benefit.  We can enable low-cost infrastructure  Enhance all of the existing projects  Enable new projects that were previously intractable

6. Population (in millions) >$20,000 $2,000–20,000 <$2,000 Purchasing power parity (in U.S. dollars) There is a market, too… 100 4,000 emerging ‘mass’ markets adjacent markets 2,000 source: Harvard Business Review © 2002

7. ICT4B Information and Communication Technology for Billions

8. Big Picture  Enhance and enable IT projects:  Novel technology (direct attack)  Novel deployment/support  Support for semi- and illiterate users  Two real-world deployments (validate)  Great partners

9. Novel Technology  Device cost: 10-100 times reduction  Infrastructure cost: 10-100 times reduction  Device power: 10-100 times lower  Speech recognition for obscure languages and dialects

10. Three Layer Architecture  Devices  1-70 users each, $1-10  Short range wireless  Proxies (basestations)  100-1000 users, $200, < $1/user  Mixed wired, wireless, satellite  Transient storage  Data Centers  >1M users, < $0.10 / user  Full power, networking, persistent storage

11. Devices  Co-Design Devices/Infrastructure  => 20-40x lower cost  Enables more functionality  Storage, processing, human analysis  Longer battery life  Novel low-cost flexible displays  10-50x cheaper, more robust  Printed using an inkjet process  System on a chip => $1-5 per device  Looking at 1mW per device (with radio!)

12. Data is in the Infrastructure  Manage persistent state in the infrastructure  Can lose/rent the device  Enables social science research (with privacy)  Enables group state  group calendars and news

13. Decouple Apps & Devices  Remote reprogramming over the network (authenticated)  Can upgrade/add services without changing the device!  Devices last longer  Devices increase in usefulness  Enables flexibility and research

14. Low-cost Infrastructure  Goal: 10-100 lower cost  Key idea: intermittent networking  Most apps do not need real-time continuous communication  Asynchronous is 10-100 times cheaper  Feel: some spots are interactive, most are more like e-mail  Novel protocols, app support

15. Novel Deployment & Support  Micro-franchise model for long term  Grameen Phone  Remote management for most things  Self-contained wireless proxies with ad hoc networking  No keyboard, monitor, etc. on proxies.  Data Centers are widely shared

16. Grameen Phone Lady  Micro-franchisee  Usually with a micro- loan ($200)  Buy and manages the cell phone  Rents it out to her village (10-70 users)  Income goes up 2-3x  Pays back loan (98% !)

17. Grameen Phone Lady  Micro-franchisee  Usually with a micro- loan ($200)  Buy and manages the cell phone  Rents it out to her village (10-70 users)  Income goes up 2-3x  Pays back loan (98% !) Uganda

18. Literacy  Novel speech recognition:  Easy to train  Speaker independent  Any language or dialect  Small vocabulary (order 100 words)  A non-IT person can train the speech for her dialect  Also speech output (canned)  May do recognition on the device, or on proxy

19. Real Deployments  First one is in India (2005-6)  IIT Delhi, HP Labs India  Second outside of Asia (2006-7)  Probably Africa, Mexico or US

20. Five Application Areas  Commerce  Health  Education  Government  Location-based services  Team includes social scientists:  Stephen Weber, Isha Ray

21. Great Partners  HP  Intel  Grameen Bank  UNDP  Markle  IIT Delhi More welcome!!!

22. Summary  New approach for IT in developing regions  Novel technology, infrastructure  “direct attack” on the key challenges  Real deployments  Enable and enhance 1000s of projects worldwide  Long term: development over defense

24. Reduced overall cost  Functionality moved to the infrastructure costs far less!  Device utilization = 4%  Infrastructure servers = 80%  Ex: ISPs have 20-40 times less modems than there customers, even though every connection takes one from each...  Admin & support costs also decrease

25. Transformation Examples  Tailor content for each user & device  Noticeably faster than home PC + modem 1.2 The Remote Queue Model We introduce Remote Queues (RQ), …. 1.2 The Remote Queue Model We introduce Remote Queues (RQ), …. 65x 6.8x 10x

26. Refinement  Retrieve part of distilled object at higher quality Distilled image (by 60X) Zoom in to original resolution