Polychronis Ypodimatopoulos

1
Parallel Internets and Ultra-Local Economies

• Polychronis Ypodimatopoulos
• Viral Communications group
• MIT Media Laboratory
• CFP Bi-Annual Meeting
• San Jose
• January 2008

2
Project goal

• Organize the presence, profile and social
  interaction of humans and objects in physical
  proximity and make it accessible and useful

3
Social Networking

• Not really new The Network Nation, S. Hiltz,
  M. Turoff (Addison-Wesley, 1978, 1993)
• Others followed
• USENET
• classmates.com
• sixdegrees.com
• myspace.com
• facebook.com
• linkedin.com

4
How social really is Facebook ?

• Ypod 98 friends, 108 total messages

Colleague in MOTOROLA
friend in Greece
72 friends in MIT Where is my social
interaction?!?
friend in UK
others None of people that I interact with on
daily basis
5
How social really is Facebook ?

• Facebook usage maximizes between 9pm-12am and
  plummets between Friday afternoon and Sunday
  afternoon
• Traffic also increases during summer/winter breaks

Is Facebook really a tool for initiating social
interaction, or for merely maintaining it?
Fraction of messages sent to recipients in the
same school in 2005
Rhythms of social interaction messaging within
a massive online network, Golder, Wilkinson,
Huberman
6
Bottom-up approach to Soc. Networking

• Primary characteristic of social interaction for
  vast majority of humans?
• Location of participants to most popular social
  networking tools/platforms?
• Examples
• You can find the location of a building in a
  city, but have no idea on which side its entrance
  is
• You look for people with common interests, but
  fail to discover those sitting next to you on the
  train
• You live in large apartment building, but have no
  means of establishing social interaction with
  neighbors (other than door2door, if you dare)
• Two strangers at the airport take separate taxis
  to go to the same location, etc.

Common physical location
Virtual location
7
Bottom-up approach to Soc. Networking

• We build a mesh network of humans and objects in
  physical proximity
• Each entity participates by means of a device
  that carries a public profile about its owner
  (humans interests, location of a door, etc)
• The confederation of all profiles in the network
  yields a new type of data that is specific to the
  profiles and the location of the entities
• We can query this data to draw useful
  information, discover entities based on their
  location and help establish social interaction

8
Proposed solution Cerebro

• Suppose there is a number of users and/or objects
  in physical proximity

9
Proposed solution Cerebro

• Cerebro discovers the presence of all other
  entities and offers asymmetric information
  resolution about the layout of the network
  (boosts scalability)

10
Proposed solution Cerebro

• A profile is stored at each entity and it is
  accessible throughout the network
• We have organized data that was previously
  unavailable into useful and accessible information

11
Proposed solution Cerebro

• Multiple mesh networks tunneled together form a
  Parallel Internet

12
Assumptions

• User carries some WiFi device that is (almost)
  always on
• User regularly updates her profile to match her
  day-to-day needs/mood/interests

13
The Result

• On the Street
• Potential clients are literally declaring
  products/services they need
• Discover your peers, combine your (buying) power
• Express any of multiple identities based on
  different contexts
• Communication in emergency situations

14
The Result

• At Home
• Discover neighbors with similar interests, share
  playlists, integrate into TV set
• Integrate into alarm systems and communicate
  emergency situations to neighbors
• At the Workplace
• Organize and search for know-how by physical
  location

15
The Challenges

• Extreme scalability
• Efficient search for information
• Reflect users social norms onto the behavior of
  the device
• Security

16
Progress so far

• Achieving scalability on a diet Connected 100
  nodes in mesh network using a single frame per
  node, per 10 seconds (15kb/sec in the worst case)
• Portability Cerebro runs on x86, OLPC XO, Nokia
  N800 and ARM-based embedded computers (python)

17
Next steps

• Introduce a multi-radio device and demonstrate
  communication symmetry between humans and
  objects
• Discover some object (your door at office, your
  car or your scooter)
• Express one of your identities (by means of RFID)
• Establish communication and exchange profiles
• Get statistics from your home entrance (whos
  inside?)
• Sync your MP3s with your car/scooter
• Customize your car/scooter settings

18

• Questions?