Fujitsu Laboratories of America Technology Symposium 2015 Expectations for Intelligent Computing Tango Matsumoto CTO & CIO FUJITSU LIMITED
Outline What s going on with AI in Fujitsu? Where can we apply AI technology? Challenge in AI Technology Development 1
A Challenge using Machine Learning Computer became able to win professional Shogi* player. * Japanese chess Year Score Computer 2012 1-0 2013 3-1 2014 4-1 2015 2-3 10-5 Professional 2
Machine Learning is key with Shogi Checker size Chess piece Kinds Number Technology to beat professional players Chess 8 x 8 6 16 x 2 Knowledge Acquisition from Human Expert < 1000 parameters Shogi 9 x 9 8 20 x 2 (Reusable) Machine Learning from 60,000 past match data > 100,000,000 parameters 3
Artificial Brain Project* Goal: Cross the threshold required for admission to University of Tokyo by 2021. Fujitsu in charge of Mathematics. 60(σ=1) Automatically solved two out of four math questions. 15.9% Deviation value of 60 Input Test problem Procedure Natural language processing Formula interpretation Knowledge inference Computer algebra Output An answer by the computer * National Institute of Informatics project 4
Analytics in IoT Era Multiple dimension time series data Technology to find hidden relations among big data by two step machine learning Application; Fault prognosis for equipment in factories or plants. Multiple Dimensions Step 1 Machine learning with normal data Step 2 Current Voltage Temperature Pressure Machine learning with abnormal pattern Normal status Fault Time line Possible abnormal patterns Abnormal patterns before having fault 5
Where can we apply AI technology? 6
Artificial Intelligence Landscape Individual Enterprise Industry & Public Sector Knowledge and thinking Decision making support Communication Movement / Activity Handicapped / Aged Tough physical work Sports Skills Human interface Emotion Estimation Augmented Reality Phenomena estimation Decision making support Marketing Security / authentication Call center QA system Education Medical Autonomous Driving Environmental / Energy Traffics / Mobility Social Infrastructure Digital contents Retail Finance Agriculture Sports Education Dealing with disasters 7
QA System Factoid type Questions to ask What, When, Where, or Who What Event, Object, Number, When Date, Year, Where Location, Country, Who Person s name, (objective facts) How Why Others Non-factoid type Questions to ask Why or How Procedure, Process, Action, Suggestion, Proposal, Cause, Reason, Experience, Search method Reasoning method 8
Recognition of Intention, Emotion, and Status Enable more comfortable or satisfying services. Enable support to prevent from undesirable failure or accident. Irritated AI Satisfied A look Eyes Movement Way of speaking Intention Emotion Status Customization Personalization Fail-safe Big data analysis by Deep Learning 9
High Value-Added Food Lettuce raised in sanitized Clean Room 1. Delicious! 2. Eatable without wash! 3. Fresh for more than a month! 4. Low potassium! Value-added Delicious Long time preservation High yield Food Safety Nutrition control Health Germfree pack Clean Room Find raise conditions by Machine Learning 10
Sustainable Maintenance of equipment or Infrastructure Fault prognosis from series of data Prevention from accidents Effective recovery from accidents Equipment Production line Plant and Equipment Infrastructure 11
Disaster Prevention, Damage Reduction Disaster prediction, Quick effective countermeasure against disaster by AI. Earthquake Flood Mudslide Volcanic eruption Forest fire 12
Challenge in AI Technology Development 13
ICT and AI are in Products and Services Services will be changed by AI Medical Electronic medical Record, Medical prescription Agriculture Sales/ Distribution management Manufacturing Production management Artificial Intelligence (AI) Health analysis Harvest Value-added vegetables Usage analysis Human Centric Innovation 14
Fair use of AI Technology Problems in AI implementation Goodness Reliability of used data Security & privacy Social receptivity Governance for fairness 15 15
Challenge to find Inspiration mechanism Joint research with Riken (Japan s Institute of Physical and Chemical Research) Fusion of brain science and computer science for advanced prediction. Perception of board patterns Intuitive best next-move generation Neural Basis of Intuitive Best Next-Move Generation in Board Game Experts Science, 21 Jan. 2011, Volume 331 Issue 6015. 16
Together, we can make it happen! 17
18 Copyright 2010 FUJITSU LIMITED
Cautionary Statement These presentation materials and other information on our meeting may contain forward-looking statements that are based on management s current views and assumptions and involve known and unknown risks and uncertainties that could cause actual results, performance or events to differ materially from those expressed or implied in such statements. Words such as anticipates, believes, expects, estimates, intends, plans, projects, and similar expressions which indicate future events and trends identify forwardlooking statements. Actual results may differ materially from those projected or implied in the forward-looking statements due to, without limitation, the following factors: general economic and market conditions in the major geographic markets for Fujitsu s services and products, which are the United States, EU, Japan and elsewhere in Asia, particularly as such conditions may effect customer spending; rapid technological change, fluctuations in customer demand and intensifying price competition in the IT, telecommunications, and microelectronics markets in which Fujitsu competes; Fujitsu s ability to dispose of non-core businesses and related assets through strategic alliances and sales on commercially reasonable terms, and the effect of realization of losses which may result from such transactions; uncertainty as to Fujitsu s access to, or protection for, certain intellectual property rights; uncertainty as to the performance of Fujitsu s strategic business partners; declines in the market prices of Japanese and foreign equity securities held by Fujitsu which could cause Fujitsu to recognize significant losses in the value of its holdings and require Fujitsu to make significant additional contributions to its pension funds in order to make up shortfalls in minimum reserve requirements resulting from such declines; poor operating results, inability to access financing on commercially reasonable terms, insolvency or bankruptcy of Fujitsu s customers, any of which factors could adversely affect or preclude these customers ability to timely pay accounts receivables owed to Fujitsu; and fluctuations in rates of exchange for the yen and other currencies in which Fujitsu makes significant sales or in which Fujitsu s assets and liabilities are denominated, particularly between the yen and the British pound and U.S. dollar, respectively.