Document Sample
Lifeline Powered By Docstoc
					Applications of Intelligent Systems
and Robotics in Service of Society

                     Raj Reddy
              Carnegie Mellon University
                        Jan 9, 2007

       Keynote Speech at IJCAI 2007, Hyderabad, India

         Outline of the Talk

 Needs    of Developing Economies
      Accessto Knowledge, Education and
      healthcare, etc.
3  Minute Introduction to AI: What is it and
  how it can help
 The role of AI in enabling
      access to knowledge and knowhow
         access to libraries
      access to education and learning
      access to health care
 Unfinished    research agenda of AI
             Needs of the People

with Per Capita Income of Less Than $1 a Day
    Access to entertainment
        watch any movie, TV show when desired
    Telemedicine
        providing links to doctors and treatment at a distance
    Access to information
        about hygiene and safe water, helping to reduce infant
    Life-long learning
         independent of the limitations of language, distance, age and
         physical disabilities
    Price discovery
    Marketing assistance
        using eBay like auction exchanges
    Find jobs
        e.g.

                    They need AI and IT
            but not Word, Excel and Powerpoint

Barriers to Entry: The Digital Divide
    Connectivity Divide
       Access to free Internet for basic services?
    Computer Access Divide
       Accessibility: Less than 5 minute walk?
       Affordability: Costing less than a cup of coffee per day?
    Digital Literacy Divide
       Language Divide
       Literacy Divide
    Content Divide
       Access to information and knowledge
       Access to health care
       Access to education and learning
       Access to jobs
       Access to entertainment
       Access to improved quality of life

      A 3-Minute Introduction to AI
   What is it and how it can help
   review why the world’s poor have more to gain
    in relative terms by the effective use of the IT
    and AI technology

Artificial Intelligence attempts to make
computers do things which would require
intelligence in people, i.e. any activity which
requires the use the human brain

    A Historical View of Advances in AI

 1950s: Theorem Proving; Chess
 1960s: Problem Solving;
         Language: Understand; Question Answering
 1970s: Speech; Vision; Expert Systems
 1980s: Robotics; Knowledge Based Systems
 1990s: Language Translation; Search
 2000s: Systems that Learn with Experience

    Some Application Domains
   Web Search : Google, Yahoo, MSN
   Intelligent car
   Financial planning
   Manufacturing control
   System diagnosis
   NL communicator
   Writing assistant
   Knowledge-based simulation
   Games
   Household robot

    Requirements for Intelligence

 Learn from experience
 Exploit vast amounts of knowledge
 Exhibit Goal Directed Behavior
 Tolerate error and ambiguity in input
 Communicate with natural language
 Operate in real time, and
 Use symbols (and abstractions)

    AI Problem Domains & Attributes
                   Knowledge   Data     Response
                     Content    Rate        Time
Puzzles              Poor       Low        Hours
Theorem Proving
Expert Systems
Natural Language
Motor Processes
Vision                Rich       High      Real Time

       Lessons from AI Experiments
   Bounded Rationality implies Opportunistic
   An Expert becomes a World Class Expert only
    after spending at least 15 years of intensive
    practice and knows 70,000+20,000 patterns
   Search Compensates for Lack of Knowledge
   Knowledge Compensates for Lack of Search
   A Physical Symbol System is Necessary and
    Sufficient for Intelligent Action

              How Can AI Help?

   Intelligent Systems in support of
     Access to Knowledge and Knowhow
     Learning and Education
     Health
     Robotics for Accident Avoiding Cars,
      Landmine Detection, and Disaster Recovery

   Enabling Access to
Knowledge and Information
            Village Google:

Access to Knowledge for Use in a Village
   Access to Essential Information and Advice
       Medical, Agriculture, FAQ indexed and searchable
       Interactive access to Doctors, Rescue Personnel
   Lifelong Learning and Education
   Agricultural Information
       Price discovery, crop disease information, weather
   Access to Markets and Jobs
   Disaster Relief and Management
   Access to Newspapers, Radio and TV
   Entertainment and Amusement
   Communications
       Video Phone, IP Telephone, Instant Messaging
       Video Email, Voice Email, Text Email

    The Vision of a Global Knowledge Network
   Create a Knowledge Network that connects experts to the
    people who need help, e.g., farmers in villages
   End-users interact at Village Knowledge Centers
        Equipped with a networked computer and basic A/V equipment
        Staffed by a Knowledge Officer
             Humans are intrinsic to Knowledge Networks
              (raw information  knowledge!)
   Domain experts provide answers to previously unanswered
   Answers converted into an “encyclopedia-on-demand” video
    documentary at higher-level centers centers and dubbed into
    local languages in each country
   Also available for direct access browsing by literate and
    networked users
                                    System Overview

Multi-level Information Flow - An example scenario

An illiterate farmer goes to a   The KO retrieves        For the remaining 10 - 20%   100s of domain experts
Village Knowledge Officer        answer from local       of the time the KO puts up   populate the databases,
(with a computer connected       Multilingual database   the question to a higher     both as part of their jobs
to FAO multimedia                within minutes 80 -     level office and gets an     and as volunteers (say, 2
database) and asks a             90% of the time         answer back, typically       questions per week)
question in his or her local                             in less than 24 hrs

   Hierarchical structure spanning districts, regions, countries, etc.
   Outside experts interact with higher level Knowledge Officers
   Builds up an ever-increasing multimedia database
         Can provide static (e.g., best-practices) as well as dynamic (e.g.,
          weather, prices, etc.) information
   Innovative mechanisms and processes for information
    digitization, exchange, analysis, and dissemination

Knowledge officers and Domain Experts
                                          World      Knowledge Management
                                                     & Coordination (global)

                                             Knowledge Management
                                             & Coordination (national level)
                                       Verification of Query-Answer Relevance
                                       And RFP to domain experts


                                 Translation, Information Retrieval


                   AV data collection,
                   Transliteration and Transcription
                   Information Retrieval
                                                             Domain experts:
                                                             Volunteer to
                                                             answer at least 2 questions a week
                                                             (or part of job responsibility)

                             Roles of Knowledge Officers
            Village                             District                  Region/Nation                          (sub)continent                      Global

      3,000 people                       300,000 people                       30M people                           0.3B people              3 Billion people

Transcription (and possibly               Translation and                    Verification &                   Knowledge Management          Knowledge Analysis
     Transliteration)                  Information Retrieval               RFP from Experts                       & Coordination              and Inference
Records question of the end-user     Enters translation of questions.    Picks questions of critical nature   Same as next level up, but   Brings experts to where their
in audio-video format. Enters text                                       and validates the answer             with the range of analyses   knowledge is needed.
transcription of the question.       Searches multilingual database      provided at lower level              broadened to the
                                     for answer                                                               region/subcontinent level    Mobilization of resources
Searches local language                                                  If critical or unanswered                                         towards their need.
database for answer                  Sends answer after translation to   question, puts up request to
                                     lower level                         experts even if not paid for                                      Identifies and triggers
Need not be knowledgeable in                                             by end-user                                                       initiatives to control
English.                             If question not among FAQs or                                                                         “epidemic”-like problems
                                     automated system, sends to
                                     higher level

     (All numbers shown are for rural, developing country populations = beneficiaries)
            The AI Challenges in

    Creating a Global Knowledge Network
   Farmers typically not able to tap in to existing
       Often illiterate
       Rarely have relevant information or even communications
   Today’s Internet and existing databases/portals are
    primarily intended for users literate in English and can
    synthesize their solutions from multiple sources

                 Internet Bill of Rights
                         Jaime Carbonell, 1994

   Get the right information
       e.g. search engines
   To the right people
       e.g. categorizing, routing
   At the right time
       e.g. Just-in-Time (task modeling, planning)
   In the right language
       e.g. machine translation
   With the right level of detail
       e.g. summarization
   In the right medium
       e.g. access to information in non-textual media

          Relevant Technologies

   “…right information”      search engines
   “…right people”           classification, routing
   “…right time”             anticipatory analysis
   “…right language”         machine translation
 “…right level of detail”  summarization
 “…right medium”           speech input and output

“…right information”
   Search Engines

                       The Right Information
   Right Information from future Search Engines
       How to go beyond just “relevance to query” (all) and “popularity”
   Eliminate massive redundancy e.g. “web-based email”
       Should not result in
            multiple links to different yahoo sites promoting their email, or even non-
             Yahoo sites discussing just Yahoo-email.
       Should result in
             a link to Yahoo email, one to MSN email, one to Gmail, one that
             compares them, etc.
   First show trusted info sources and user-community-vetted
       At least for important info (medical, financial, educational, …), I want
        to trust what I read, e.g.,
            For new medical treatments
                First info from hospitals, medical schools, the AMA, medical publications, etc.
                 , and
                NOT from Joe Shmo’s quack practice page or from the National Enquirer.

              Beyond Pure Relevance in IR
 Current Information Retrieval Technology Only
  Maximizes Relevance to Query
 What about information novelty, timeliness,
  appropriateness, validity, comprehensibility, density,
 Novelty is approximated by non-redundancy!
       we really want to maximize: relevance to the query, given
        the user profile and interaction history,
            P(U(f i , ..., f n ) | Q & {C} & U & H)
             where Q = query, {C} = collection set,
             U = user profile, H = interaction history
       ...but we don’t yet know how. Darn.

Maximal Marginal Relevance vs.
 Standard Information Retrieval

       Standard IR


“…right people”
Text Categorization
                         The Right People

   User-focused search is key
        If a 7-year old is working on a school project
             taking good care of one’s heart and types in “heart care”, she will want links
              to pages like
                  “You and your friendly heart”,
                  “Tips for taking good care of your heart”,
                  “Intro to how the heart works” etc.
                  NOT the latest New England Journal of Medicine article on “Cardiological
                   implications of immuo-active proteases”.
        If a cardiologist issues the query, exactly the opposite is desired
        Search engines must know their users better, and the user tasks
   Social affiliation groups for search and for automatically categorizing,
    prioritizing and routing incoming info or search results. New machine
    learning technology allows for scalable high-accuracy hierarchical
        Family group
        Organization group
        Country group
        Disaster affected group
        Stockholder group
               Text Categorization

Assign labels to each document or web-page
 Labels may be topics such as Yahoo-categories
       finance, sports, NewsWorldAsiaBusiness
   Labels may be genres
       editorials, movie-reviews, news
   Labels may be routing codes
       send to marketing, send to customer service
               Text Categorization

 Manual assignment
       as in Yahoo
   Hand-coded rules
       as in Reuters
   Machine Learning (dominant paradigm)
       Words in text become predictors
       Category labels become “to be predicted”
       Predictor-feature reduction (SVD, 2, …)
       Apply any inductive method: kNN, NB, DT,…

    “…right timeframe”
Just-in-Time - no sooner or later
             Just in Time Information

   Get the information to user exactly when it is
       Immediately when the information is requested
       Prepositioned if it requires time to fetch & download
        (eg HDTV video)
            requires anticipatory analysis and pre-fetching

   How about “push technology” for, e.g. stock
    alerts, reminders, breaking news?
       Depends on user activity:
            Sleeping or Don’t Disturb or in Meeting  wait your chance
            Reading email  now if info is urgent, later otherwise
            Group info before delivering (e.g. show 3 stock alerts

“…right language”

    Access to Multilingual Information
   Language Identification (from text, speech, handwriting)
   Trans-lingual retrieval (query in 1 language, results in
    multiple languages)
       Requires more than query-word out-of-context translation (see
        Carbonell et al 1997 IJCAI paper) to do it well
   Full translation (e.g. of web page, of search results
    snippets, …)
       General reading quality (as targeted now)
       Focused on getting entities right (who, what, where, when
   Partial on-demand translation
       Reading assistant: translation in context while reading an original
        document, by highlighting unfamiliar words, phrases, passages.
       On-demand Text to Speech
   Transliteration
        “…in the Right Language”

   Knowledge-Engineered MT
       Transfer rule MT (commercial systems)
       High-Accuracy Interlingual MT (domain focused)
   Parallel Corpus-Trainable MT
       Statistical MT (noisy channel, exponential models)
       Example-Based MT (generalized G-EBMT)
       Transfer-rule learning MT (corpus & informants)
   Multi-Engine MT
       Omnivorous approach: combines the above to
        maximize coverage & minimize errors

“…right level of detail”

                 Right Level of Detail
   Automate summarization with hyperlink one-click
    drilldown on user selected section(s).
   Purpose Driven: summaries are in service of an
    information need, not one-size fits all (as in Shaom’s
    outline and the DUC NIST evaluations)
       EXAMPLE: A summary of a 650-page clinical study can focus on
          effectiveness of the new drug for target disease
          methodology of the study (control group, statistical rigor,…)
          deleterious side effects if any
          target population of study (e.g. acne-suffering teens, not eczema
           suffering adults ….depending on the user’s task or information
        Information Structuring and

   Hierarchical multi-level pre-computed summary
    structure, or on-the-fly drilldown expansion of info.
       Headline <20 words
       Abstract 1% or 1 page
       Summary 5-10% or 10 pages
       Document        100%
 Scope of Summary
       Single big document (e.g. big clinical study)
       Tight cluster of search results (e.g. vivisimo)
       Related set of clusters (e.g. conflicting opinions on how to cope
        with Iran’s nuclear capabilities)
       Focused area of knowledge (e.g. What’s known about Pluto?
        Lycos has good project in this via Hotbot)
       Specific kinds of commonly asked information(e.g. synthesize a
        bio on person X from any web-accessible info)
      Document Summarization

                    Types of Summaries

        Task                 Query-relevant          Query-free
                                (focused)             (generic)
    INDICATIVE             Filter search engine    Short abstracts
    for Filtering                  results
(Do I read further?)
   CONTENTFUL            Solve problems for busy      Executive
for reading in lieu of        professionals          summaries
       full doc

         “…right medium”
Finding information in Non-textual Media

          Indexing and Searching
        Non-textual (Analog) Content
 Speech  text (speech recognition)
 Text  speech
       TTS: FESTVOX by far most popular high-quality
 Handwriting  text (handwriting recognition)
 Printed text  electronic text (OCR)
 Picture  caption key words (automatically) for
  indexing and searching
 Diagram, tables, graphs, maps  caption key
  words (automatically)

  AI and Access to Libraries
The Million Book Digital Library Project
               One Step at a Time…

   Million Book DL
       Only about 1% of all the world’s books
            Harvard University    12M
            Library of Congress   30M
            OCLC catalog 42M
            All Multilingual Books ~100M

   At the rate of digitization of the last decade it
    would take a 100 years!
         Million Book Project: Issues
   Time
       At one page per second (20,000 pages per day
        shift), it will take 100 years (200 working days per
        year) to scan a million books of 400 pages each
   Cost
       100M books at US$100 per book would coat $10B
       Even in India and China the cost will be $1B
       The annual cost is currently expected to be close
        $10M per year with support from US, India and
   Selection
       Selection of appropriate books for scanning is time
        consuming and expensive
Million Book Project: Issues (cont)
   Logistics
       Each containers hold 10,000 to 20,000 books.
        Shipping and handling costs about $10,000

   Meta Data
       Accessing and/or creating Meta data requires
        professionals trained in Library science

   Optical Character Recognition Technology
       Essential for searching, translation and
       Many languages don’t have OCR
        Million Book Project: Status
   18 Centers in India
   22 centers in China
   1 Center in Egypt
   15 Centers in Poland
   Planned : Australia
   Over 1,400,000 books scanned
       Over 250,000+ accessible on the web
Title       Rig Veda
Author      Pandit Sriram Sharma Acharya
Language    Sanskrit
Subject     Philosophy
Publisher   Sanskriti Sansthan Bareli
Abstract    Rig Veda is the oldest of the
            Vedas. The Rig Veda is the
            oldest book in Sanskrit or any
            Indo-European language. Many
            great Yogis and scholars who
            have understood the
            astronomical references in the
            hymns, date the Rig Veda as
            before 4000 B.C., perhaps as
            early as 12,000. Modern
            western scholars date it around
            1500 B.C., though recent
            archaeological finds in India
            (like Dwaraka) now appear to
            require a much earlier date
Title       Elementary Treatise on the
            Wave-Theory of Light
Author      Humphery Lloyd, D.D, D.C.L
Language    English
Subject     Physics
Publisher   Longmans, Green & Co
Year        1873
Abstract    This book deals with the
            various aspects of the wave
            theory of light. It is a critical
            work which contains an
            analytical discussion of the
            most recent researches in
            Optics. It presents a clear and
            connected view of the
Title       Mudalayiram Mulamum
Author      Periya Jeeyar
Language    Tamil
Subject     Religion
Publisher   Sri Vaishnava Sampirathaya
            Sanjeevikiri Sabayai
Year        1909
Abstract    This volume is written in Tamil.
            It provides a detailed account
            of the origin of Vaishnava and
            is written by Periya Jeeyar. .
Title       Gulzar-A-Badesha
Author      Khader Badesha
Language    Urdu
Subject     Literature
Publisher   Namipress, Chennai
Year        1919
Abstract    Literature
Title       Jawahar Ali Joyviyah
Author      Dr.Ilyas lomas
Language    Arabic
Subject     Metrology
Publisher   Bakri and Issa
Year        1876
Abstract    It is a book on Metrology, a
            study of measurements
Title       Structure Des Molecules
Author      Victor Henri
Language    French
Subject     Chemistry
Publisher   Taylor and Francis
Year        1925
Abstract    This is a unique book that
            explicates, in detail, the
            structure of molecules and
            touches upon certain specific
            characteristics of molecules
            with particular reference to
            Million Book Project:
           AI Research Challenges
 Multilingual Information Retrieval
 Translation
 Summarization
 Reading Assistant using Multi Lingual
  Speech Synthesis and Translation (e.g. for
  news paper DL)
 Easy to use interfaces for Billions
 Providing Access to Billions everyday
     Distributed   Cached Servers in every region

AI and Education
 Intermediate Examination 2006

                  Urban – Rural Divide



Passing   First division   More than 75%   More than 90%   First division-   First division-
                                                           maximum of a      minimum of a
                                                               district          district

                                   Rural   Urban
Intermediate Examination 2006

Differences in Performance of Different
     Social Groups – Percent Failing

       43                                 43



  FC   BC    SC   ST   Muslim   Others   Total
Intermediate Examination 2006

 Differences in Performance of Different
              Social Groups



   40                                       4

        20   2                                              23
                            2              17
             10             8

  FC    BC   SC             ST           Muslim   Others   Total

             75 % or more        more than 90 %
Performance in EAMCET 2006

                            Rural Urban Divide
 Percent share


                  Avg. of Math+Sci EAMCET rank less EAMCET rank less EAMCET rank less
                 greater than 94.5%   than 5,000      than 10,000      than 50,000

                                             Rural   Urban

     Large Variation in School Quality
   No. of schools where NOT a SINGLE
    student got more than 75% marks and
    more than 50% of all taking exam failed
        360 in 2004, and
        965 in 2006

   Intensity of problem is almost twice in
    rural areas compared to urban areas

     Large Variation in College Quality
   Even bright fail!
         1345 students who got more than 90% in Math in SSC failed in
          either math A or B in year I or year II
         Of these 1345, 222 had >90% in two subjects and 53 in three
   253 colleges where failing rate is more than 75%
   239 colleges where not a single student gets more than
   829 colleges where less than 5% students passing with
    more than 75% (state avg. is 22%)
   Intensity of problem is almost twice for colleges in rural
    areas compared to colleges in urban areas

    Problems with Current System
   Focus on national best with consequent neglect of local
       Urban students with access to tuition and coaching get the
        highest ranks in national tests
   Schools in remote villages
       Lack of quality teachers
       No coaching centers
       Deprived of competitive atmosphere
   No system to nurture talent who do best in such difficult
       Financial issues often prohibit the brightest rural students from
        attending the best universities

Problems with Current System (Cont)
   Lack access to quality colleges
   Lack proper guidance, motivation and peer
   Inadequate support from families
   Poverty prevents access to coaching classes,
    tutoring etc
   Poverty compels them to seek work to for
    livelihood rather than proceed to college
    essential for reaching their full potential

           Current System
Admission to Engineering and Medicine
   Coaching for 11th and 12th (costs 60K to
    120/240K), Kota, Hyderabad, Delhi,
   Unaffordable to many
   Teaching to test
     Not   broad education
   Revised pattern of JEE seems not to
    diminish the importance of coaching

             During Formative Years
   Right guidance and environment during formative
   This is what famous mathematician Hardy says
    about mathematics genius Srinivas Ramanujan
      The years between eighteen and twenty-five are the critical years
      in the mathematician’s career and that the real tragedy is not that
      Ramanujan died early, but during these years his genius was
      misdirected, sidetracked, and to some extent even distorted

    Problems with Current System
 Wastage of precious time
    commuting (lot of time in to-and-fro, may be
     1-4 hours a day)
    only two semesters in a year
 Lack focus on development of soft skills, a key
  to success in today’s highly competitive job
 Imperfect credit market for higher secondary
     Have you heard of bank loan for “coaching classes”   for 11th
     and 12th, JEE, EMCET, AIEEE

                 How AI can Help?

Creating a New Affirmative Action Plan For The
  Socially Disadvantaged?
 Data Mining: Local Best instead of National Best
 Intelligent Tutoring Systems (AI Meets Cognitive
  Science) : Variable Duration Learning
       Online Reading Tutors
       Online Math Tutors
   Intelligent Monitoring Systems
     Early Detection of Promising Students and Problem
      Students thru Progress Monitoring
     Process Improvement
 AI and Development of Soft Skills

 Soft skills have become key to success in
  today’s highly competitive job market
 Develop Intelligent Tutoring Systems for:
     Communication skills/language proficiency
     Interpersonal Interaction and Negotiation
     Personality traits/sociability
     Teamwork
     Work ethic
     Courtesy
     Self-discipline, self-esteem and self-confidence
     Presentation skills

AI and Healthcare
           PCtvt UI Design

    for Use by Illiterate Persons
An Illiterate person needs a more
 powerful PC than a PhD!
  If not e-mail, use voice-mail
  Replace Text Help by Video Help
Radically simple design
  One minute learning time
  Two click model
  Three modes of communication: Video,
   Audio and Text
    Both Synchronous and Asynchronous
All-Iconic interfaces
Multiple input modalities
  TV-remote, Speech I/O, Keyboard, Mouse
   or Cell phone

                       AI and eLearning
 Give man a fish and you will feed him for a day. Teach
  man to fish and you will feed him for life. (Old Chinese
  Proverb -- Lao Tzu)
 How to teach an illiterate villager who has never seen a
  computer to effectively use PCtvt?
       Self-evident, intuitive interfaces
          Two clicks to most applications
          Learning time – less than five minutes to happiness
       Just in Time learning
          Immersive Interactive Simulated Environments
          Short video clips: Instant access to information through vast video
           digital libraries in local languages
       Interactive Problem Solving
            Intensive programs for educating the local expert, the Village
             Information Officer
            Teach the Teacher Programs

   A Call to Action to
AI Researchers In India
     India Has 21 Official Languages!

       We need to Break the Language Barrier!

•   Language barriers can significantly slow
    down the economic growth
•   Globalization requires cross-border and
    cross-language communication
•   Eliminate cultural and social barriers
•   Access to rare (and potentially beneficial)
    knowledge requires eliminating the
    language divide
•   Preservation of minority languages, cultures
    and heritage

    Unfinished Research Agenda for AI

 spoken language understanding,
 dialog modeling,
 multimedia synthesis and language
 multi-lingual indexing and retrieval,
 language translation, and
 summarization.
                  Next Steps

 Create technologies and solutions for
  overcoming the language barrier
 Create toolkits for rapid acquisition of new
  language capabilities
     Character codes, optical character recognition,
     speech recognition, speech synthesis,
     translation, search engines, text mining,
     summarization, language tutoring, etc.
 Capture data, information and knowledge
  from masses
 Make fundamental advances in language
  processing algorithms, e.g.,
     Deal with 1000 times more data
     Conceptual advance in semantic   information
        The Educational Plan

 Training a generation of researchers to
  explore many techniques in many
 Training innovators and entrepreneurs in
  applications of language technology
 Training scholars in each country to be
  expert in language technology
 Training individuals in foreign languages
  and cultures
           The Research Plan

 Analogy to Human Genome Project
 Meticulous core-science based fundamentals
 Researcher toolkits for known methodologies
 Architecture supporting diversity of
 Long planning horizon to support
  development of novel and radical approaches
 Quantitative evaluation against a standard of
  steadily accumulating improvements in
           Impact and Benefits

 greater participation in global economy
 preserve local languages and cultures
 promote greater communication and
  understanding among states and
 With over 100 orphan languages, each
  country of the world needs these tools in
  its own enlightened self interest
     International focus and multinational
     involvement will establish India as a world
     leader in this important technology

 As we enter the Second 50 Years AI R&D, we
  need to ask how our work can help Society at
  large and People at the bottom of the pyramid in
 Proactive Development of Intelligent Systems for
     Access to Knowledge and Know how
     Learning and Education
     Health
     Robotics for
           Accident Avoiding Cars
           Landmine Detection, and
           Disaster Rescue and Recovery

Shared By:
jianghongl jianghongl http://