the cyc foundation

Introducing: The Cyc Foundation April 13, 2006 April 13, 2006 1 Motivations Wikimedia Foundation: “Imagine a world in which every single person is given free access to the sum of all human knowledge. That's what we're doing.” Cyc Foundation: “Imagine a world in which every single person is given free access to programs that reason with the sum of all human knowledge. That's what we're doing.” April 13, 2006 2 Topic Map – Top Level April 13, 2006 3 Cyc Reasoning System Knowledge Users Knowledge Authors Knowledge Entry Tools User Interface (with Natural Language Dialog) Other Applications Cyc API Other KBs Cyc Reasoning Modules Cyc Ontology & Knowledge Base Interface to External Data Sources External Data Sources Data Bases Web Pages Text Sources Help Find Information by Inference (+KB) Query: “Someone happy” Caption: “A man watching his daughter take her first step” April 13, 2006 5 Help Find Information by Inference (+KB) Query: “Someone happy” (x) (feelsEmotion x Happiness Positive) Caption: “A man watching his daughter take her first step” (x,y) (and (father x y) (gender x Female) (sees x y) (walking April 13, 2006 6 Help Find Information by Inference (+KB) (x) (feelsEmotion x Happiness Positive) (implies (and (isa ?BIG-EVENT HumanLifecycleMilestone) (doneBy ?BIG-EVENT ?CHILD) (sees ?PARENT ?BIG-EVENT) (children ?PARENT ?CHILD)) (holdsIn ?BIG-EVENT (feelsEmotionTypeAtLevel ?PARENT (PositiveAmountFn Pride)))) ... ... (x,y) (and (father x y) (gender x Female) (sees x y) (walking April 13, 2006 7 ALICE • • • • • • • • • • JUDGE11[9:26:0]I WILL ASK YOU SOME QUESTIONS PROGRAM[9:26:0]Do you think your plan will succeed? JUDGE11[9:26:16]WHAT IS THE COLOR OF A BLUE CAR PROGRAM[9:26:16]WHAT COLOR IS a blue car JUDGE11[9:26:35]WHERE IS SUE’S NOSE WHEN SUE IS IN HER HOUSE PROGRAM[9:26:35]Where it belongs. Try searching the World Wide Web. JUDGE11[9:26:49]WHAT HAPPENS TO AN ICE CUBE IN A HOT DRINK PROGRAM[9:26:49]Try it. JUDGE11[9:27:25]HOW IS THE FATHER OF ANDY S MOTHER RELATED TO ANDY PROGRAM[9:27:25]Fine as far as I know. Natural Language Understanding requires having lots of knowledge 1.The pen is in the box. The box is in the pen. 2. The police watched the demonstrators… …because they feared violence. …because they advocated violence. System: Ready for input. User: what do you know about anthrax? System: By “anthrax” do you mean (1) the heavy metal band or (2) bacteria? User: 2 System: Here is what I know about it: Anthrax is a species of organism. All anthrax is a bacterium. All anthrax can lethally infect mammals. Anthrax is extremely toxic for people to inhale. Iraq has biological weapons that deliver anthrax. Anthrax is opaque, tangible, microscopic, . Anthrax can be dried into spores stored indefinitely April 13, 2006 10 System: Ready for input. User: what do you know about TR1? System: I know nothing about it. What sort of thing is TR1? User: bacteria System: TR1 lethally infects_________? User: People System: By “People”, I assume you mean human beings, not People Magazine. System: Does it infect some broader category of animals (e.g., air-breathing vertebrates, chordates, placental mammals, primates, eucaryotes, etc.)? User: mammals April 13, 2006 11 System: Is TR1 similar to anthrax? User: yes System: Which of the following are true of TR1: . . . User: Al Qaida possesses some TR1. System: Ready for input. User: Does Osama Bin Laden control anything that can be the basis of weapons of mass destruction? System: Yes, some TR1. User: please explain System: Osama Bin Laden controls Al Qaida. Al Qaida posseses some TR1. TR1 is a bacteria that lethally infects people. April 13, 2006 12 April 13, 2006 13 April 13, 2006 14 April 13, 2006 15 April 13, 2006 16 April 13, 2006 17 April 13, 2006 18 April 13, 2006 19 April 13, 2006 20 April 13, 2006 21 Efficiency vs. Expressiveness C++ Continuing improvements in inference performance won’t negatively effect expressiveness. Use two cooperating languages (EL and HL) to escape the limitations of an age-old tradeoff. Efficiency PASCAL HL (heuristic level language) LISP EL (epistemological level language) First-order logic nth-order logic English, German Expressiveness April 13, 2006 22 NOW: CyN in Doom3 (2005) April 13, 2006 23 BURC: Bootstrapping Using ResearchCyc • Goal: To extend Cyc‟s knowledge base using “relationships implied to be possible, normal or commonplace in the world” • Prior work with Cyc knowledge entry has been manually oriented • How will we collect common sense without a body and manual labor…? • Read, Parse, Mine! • Proposal: Read text, Parse into a database, Extract relations between words, Propose hypothetical relations between concepts April 13, 2006 24 BURC: Basic Analogy • • • • • The Shotgun approach to the Human Genome Extract millions of fragments Knit them back together by finding commonalities Will it work for the Human Memome? James Burke: „Mr. Connections‟ Lenat’s Bootstrap Hypothesis: once Cyc reaches a certain level/scale it can help in its own development and start using NLP to augment its knowledge base April 13, 2006 25 Mining Adjective Knowledge Example • “white blouse” as factoid fragment • Hypothesis: (plausibleValueOfType Blouse mainColorOfObject WhiteColor) April 13, 2006 26 Flow of Processing BNC Data Parser 1 Frag File Parser 2 Frag File Parser 3 Frag File Parser 4 Frag File Parser 5 Frag File Merged Frag File Upper Ontology Core Theories Domain-Specific Theories Facts (Database) Extractor / DB Manager Hypothesis File Cyc/Rcyc Link Fragments DB April 13, 2006 27 (Very) Brief History of Cyc • c. 1967 – AI is used on toy problems. • c. 1977 – Expert systems reason in narrow domains. • c. 1983 – Lenat, Minsky, Feigenbaum, Kay, and others recognize need for a substrate of shared world knowledge; and realize it would take hundreds of person -years to “prime the pump”. • 1984 – Admiral Bob Inman convinces Lenat to leave Stanford and pursue this high-risk, high-payoff project (Cyc) within MCC. • 1994 – Cycorp is formed. April 13, 2006 28 April 13, 2006 29 “The driver of the power of intelligent systems is the knowledge the systems have about their universe of discourse, not the sophistication of the reasoning process the systems employ. Cyc has not only the world’s largest knowledge base, but the best represented from a technical point of view.” Ed Feigenbaum inventor of the first expert system 30 editor of the AI Handbook April 13, 2006 “People have silly reasons why computers don’t really think. The answer is we haven’t programmed them right; they just don’t have much common sense. There’s been only one large project to do something about that, that’s the famous Cyc project…”. April 13, 2006 -- Marvin Minsky 31 How has Cycorp done? • • • • • • 20 years 3 million facts and rules (hand-entered) Compelling demos Some applications (constrained by business model) The basis for much greater growth “If the right way to build an A.I. involves giving Cyc away for free, that is what we will do.” – Doug Lenat (repeatedly) – Note: Jury is out on what the “right way” is April 13, 2006 32 Cycorp: True to its Promise • OpenCyc – The entire Cyc structural ontology: FREE – 300,000 concept terms, ~2M facts and rules • ResearchCyc – – – – Equal to Full Cyc (w/ Research-only license) Source code for inference engine not released API with 18,000 functions and macros! And it doesn‟t really matter. Ability to compile in your own additions • Q: Will more be released? A: It depends. – Cycorp must financially support its own R&D. – Existing releases must result in major project benefits. April 13, 2006 33 Time for the Next Phase • Cycorp has gotten us to where we are – Representational ability – Inference ability – …and will continue (R&D leader, commercialization • The rest of the world will help get us where we are going – Breadth of content – Broad real-world diffusion The thinking that got us to where we are today is insufficient to solve the problems that exist today. To solve today's problems requires a new level of thinking. April 13, 2006 34 -- Einstein Building Cyc qua Engineering Task CYC amount known April 13, 2006 35 Building Cyc qua Engineering Task 10 years CYC 1000 years amount known April 13, 2006 36 How will we get the knowledge? Games That Matter! April 13, 2006 37 Foundation as Continuation • Are we trying to make an A.I.? – No. • Are we trying to make computers behave much more intelligently? – Yes! April 13, 2006 38 Mission (DRAFT) The Cyc Foundation has been formed as an independent not-for-profit organization to hasten the arrival of intelligent tools that will help humanity. April 13, 2006 39 Assumptions • (Currently) 9 ideas that shape strategy, objectives and policy • These may need to be validated, modified or augmented • In some cases, assumptions are followed by related policy April 13, 2006 40 Assumption #1 Long before computers are as smart as people, they will be (in some cases already have been) put to use to cure disease, address hunger problems, make important new scientific discoveries and help people work together. Smarter computers will do a better job of this. April 13, 2006 41 Assumption #2 Cycorp has developed and cared for what we believe is an important piece of the AI puzzle. They have always wanted to release it to the public, but it had to be when people could realistically develop it further on their own without in some way endangering the project. One fear was “forking”, or creating incompatible variants of the knowledge base. Cycorp and The Foundation will cooperate on 1 KB. April 13, 2006 42 Flow of Cyc Data Cyc Foundation Cycorp RCyc User Gamer / Wikipedia user April 13, 2006 Team: - Subject-matter expert - Ontologist 43 Assumption #3 The knowledge that will give computers human-like intelligence ultimately needs to be free. That's our best hope of having it put to best use. Portions of knowledge will always be held proprietary. The more shared a piece of knowledge, the greater will be the force pulling all of its representations toward freedom (to avoid the burden of maintaining a non-standard representation). April 13, 2006 44 Assumption #4 Proposed Semantic Web standards (such as those related to OWL) are an important step in the right direction, because they provide a foundation for working with meaning on the Web. The Cyc ontology will be a valuable addition, because it can act as a semantic hub, allowing us to have shared meaning. There is some concern that a top-down central ontology will dictate use of terms that may not meet a project‟s needs. We will be able to show that use of the Cyc ontology can satisfy both needs and will be a useful complement to the great work that has already been done toward the Semantic Web. April 13, 2006 45 Assumption #5 We all have something to learn. We all have something to teach. The Foundation mission will benefit from a very broad base of support, rather than the traditional rule by the technical elite. April 13, 2006 46 Assumption #6 For this effort, focused work by many will be more valuable than genius work by a few. To be most helpful, people should work together, and on tasks where they are capable of contributing successfully. (Example: don‟t go off and try to “solve the A.I. problem” by yourself.) April 13, 2006 47 Assumption #7 Regular humans can be turned off by overly technical talk that is out of place – and rightly so. We need to be inclusive in our language and in our activities in order to ensure the broadest base of support and participation. This is especially true in the Cyclify initiative. April 13, 2006 48 Assumption #8 There is no “us” and “them” • The Foundation is managed by its volunteer board and run by its volunteer members • The Foundation will start with no employees • The will be no BDFL – Benevolent Dictator for Life April 13, 2006 49 Assumption #9 Fun is mandatory! • By comparison, contributing to SETI is like cleaning your oven while you sleep. • This work will be hands-on, compelling and (hopefully) addictive. • If you‟re not having fun, find out why and fix it. April 13, 2006 50 Foundation Goals • Convert human knowledge to a form that computers can reason with – Grow the Cyc Ontology and KB Exponentially • Establish a standard vocabulary and language for representing concepts & knowledge • Support the creation of intelligent tools • Promote free and efficient knowledge transfer April 13, 2006 51 Cyclify Knowledge Collection Activities • Web Games – Validate acquired knowledge – Multiple-choice fact entry – More? • Wikipedia Linking • KR Dating Service – Wiki-based knowledge entry – A SME paired with an ontologist • WordNet Linking April 13, 2006 52 Playflow Within Cyclify RCyc User Wikipedia Data Cycorp K. Acquisiton Data RCyc Wiki Knowledge Server Game Server RCyc Team: - Subject-matter expert - Ontologist April 13, 2006 Wikipedia user Gamer 53 Status: I’m thinking of a sentence… I have 2 answers True False Don’t Know Doesn’t make sense Fibromyalgia is caused by ticks. Because I read about it on the web. Score: 24 April 13, 2006 54 Status: I think this sentence is probably not right Submitting... Thank you! Answers: 2 You agreed with: 100% I now have a better understanding of: Fibromyalgia is caused by ticks. Next Score: +2 Score: 26 April 13, 2006 55 computer (inside) Cyc Image Current Architecture computer (outside) Applet GAFs web gathered hypothesized asserted … Applet Forward rules Applet Question Server (java) KAGs SubL form, running KAG-collecting query PostGRES database Populator (java) XML file April 13, 2006 scp XML file 56 DMZ Boundary Cyc Foundation Projects • • • • • • • Nonprofit Formation (planning/budgeting/filing) Foundation Website Cyclify Fundraising Membership management Events ResearchCyc – Recommend Cyc features / functions / design – Help with ResearchCyc testing, documentation April 13, 2006 57 Budgeting • Must develop budget related to Year 1 plan • Possible areas of spending – – – – – Legal filings Server hosting W3C membership Conference attendance Fundraising April 13, 2006 58 Foundation Website • Requirements – – – – Content management features Collaboration features Out-of-the-box ease of use Free • Currently evaluating Joomla (Mambo) • Desired launch: May 15 April 13, 2006 59 Cyclify Projects • First Web Game – Develop game – Viral marketing – Add wiki linking activity • Wiki Knowledge Collection – Set up wikip.cyclify.org – Add frame for ontologizing – Feed wikip links to Web game • Back End – Design and implement PlayFlow – Submit collected knowledge to Cycorp April 13, 2006 60 Fundraising • Individual Memberships – Free membership for first 6 months for Cyclify members and ResearchCyc users? – How much? – What do you get? • Corporate Donations – Need to prepare story – Seems feasible to get donations April 13, 2006 61 What does nonprofit mean? • Cannot have investors or disburse earnings • Can have earnings, though • Revenues must come from services that are within mission • 501(c)(3)? (like Wikimedia Foundation) • Or 501(c)(6)? (like Eclipse Foundation) April 13, 2006 62 The Foundation Board of Directors Name John De Oliveira Mark Baltzegar OPEN David James OPEN OPEN OPEN TBD Sept. 2006 April 13, 2006 Position Founder and President Co-Founder and Vice President Secretary, Treasurer Board Member Board Member Board Member Board Member Board Member Role Strategy, Corp. Fundraising Strategy, Game Devel., IT Secretary, Treasurer Organizational Dynamics Standards Events, Operations Delegator Architecture, Playflow Design Oversight 63 The Foundation: Membership Project Leader, Cyclify Project Leader, ResearchCyc High Scorer (current month) High Scorer (all time) David Whitten Brad Bouldin Larry Lefkowitz Bill Jarrold Stu Baurman Kino Coursey Douglas Miles Arturo Hernandez Guyren Howe John Cabral Ben Rode Jason Azbahr Keith Wright Pierluigi Miraglia Gavin Matthews Joe Simone ~100 ResearchCyc Users YOU! April 13, 2006 64 ResearchCyc Users Government-related Government Air Force Rome Labs Houston VA Medical Center Language Computer Corporation Century Technologies ISI Daxtron Labs 21st Xerox PARC ANSER, Inc. Stone’s Throw Technologies Austin Info Systems Commercial NTT Communications Science Laboratories (Japan) Fraunhofer Institute Sapio Systems (Denmark) SRI Lockheed Martin ATLD University U of Maryland Northwestern U U of Illinois Urbana-Champaign MIT Media Lab Stanford NLP Dept. Terra Incognita Trimtab Consulting TNO-DMV (Netherlands) Microfabrica, Inc. Rensselaer AI and Reasoning Lab LBJ School of Public Affairs U of Toronto Radboud U (Netherlands) New Mexico Highlands Univ. Harvard U U of Pennsylvania Knowledge Media Institute, Open University U of Stuttgart U of Minnesota Witan International U of Hawaii Tokyo Inst. of Technology Institute for the Study Of Accelerating Change Linkoping U (Sweden) NPOs 65 April 13, 2006 How can I help? • Humans (a.k.a. common sense experts) • Programmers – Web programmers – Cyc programmers • Ontologists • Subject-matter experts • Bloggers April 13, 2006 66 Human Cyclists* • • • • • • • • • • Play the Web Game Come up with new game ideas Link Wikipedia to Cyc Learn more about Cyc Befriend an ontologist Tell a friend about Cyclify Write to a blog about Cyclify Help with viral marketing Design a logo T-Shirts: Buy one, or Create and sell them *April 13, 2006 we’re all “Cyclists” – people who interact with Cyc in one way or another. From now on, 67 Programmers • • • • • • • • Help design and build a web services interface Learn the architecture of Web Game #1 Design an add-on for the Web game Learn how to use the question server Propose a new game Help develop/support technical infrastructure Help organize documentation Help write the Cyc books – to be published by O'Reilly April 13, 2006 68 Ontologists • Identify gaps in the knowledge base • Befriend a Subject Matter Expert – Work together on a domain • Befriend a Human Cyclist – Teach one who wants to learn basic ontology skills • Help organize documentation • Help write the Cyc books April 13, 2006 69 Bloggers • Blog about Cyclify • Link to each other‟s blogs April 13, 2006 70 Timeline (Milestones) • May 15 – Launch Foundation Website • Build membership up until July 15 • June 15 – File Articles of Formation w/ Sec. Of State – First Web game in beta • July 15 – Launch Game • October – First OpenCyc build containing game data April 13, 2006 71