Published on

  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide


  1. 1. Social and ethical issue … Why have I developed the android? Hiroshi ISHIGURO Hi hi Department of Adaptive Machine Systems, Osaka University ATR Intelligent Robotics and Communications Laboratories JST ERATO Asada Synergistic Intelligence Project •Practical robot systems in the near future •Interactive robots as basic research Child android Android •120cm, 5years old •Silicon skin •Sensitive Piezo sensor (Child android) •DOF - Eys:5, Mouth:1, Neck:3 Humanlike Appearance The first contact Uncanny valley Another uncanny valley [Mori et al. ’97 ] with Prof. Itakura, Kyoto University Familiarity Familiarity Human Uncanny valley 5 years Toy robot old 1week 2weeks 18-24monthes 90 years old? Doll •Subjects from different ages (2 Similarity 100% weeks to 5 years) Moving corpse •Record subjects’ behaviors and eye motions Child android 1
  2. 2. Lateral Inhibition Hypothesis -+ -+ -+ -+ Lateral inhibition is perhaps the most fundamental operation of brain Android circuitry and integral to the operation of -+ (Adult android) all structures. 0 Response Neural network Humanlike appearance Basal ganglion Humanlike behaviors Face recognition? Humanlike perception Human identification? Growing process Uncanny valley Age-dependent uncanny valley Familiarity Familiarity 3-4 years old Toy robot Human 18-24monthes 90 years old? Human behavior recognition Uncanny valley by the sensor network Position Distributed touch sensors Position model • How to compensate the gap between androids and Distributed vision Behavior humans? Gesture recognition by vision Behavior model • Engineering knowledge is Distributed audition Dialogue Dynamic formation of Dialogue model not enough… Microphone arrays Conscious recognition of androids - The subject looks at 2 sec. - Subconscious recognition of androids - Gazing points for the uncanny android - Static android Android 77%, Human 23% Android with subconscious movements Android 30%, Human 70% Human The uncanny • Unconscious movements contribute to the reality. child android • Young people (around 20) are more sensitive than elderly people (over 30) 2
  3. 3. Eye movements while talking Android Science - Bridge between robotics and cognitive science - Mechanical eng. Development of A. I. Control sys. mechanical Human-human UP Human-robot UP Human-android UP humans Robotics Hypothesis 20 20 20 interaction UP LEFTUP RIGHT interaction UP LEFT UP RIGHT interaction UP LEFT 15 15 15 and UP RIGHT 10 10 10 Sensor tec. 5 5 5 verification RIGHT 0 LEFT RIGHT 0 RIGHT LEFT 0 LEFT human know Eveliee P1 know Repliee Q2 question know question question human think question Eveliee P1 think question Repliee Q2 Analysis and Cognitive sci. think question DOWN RIGHT DOWN LEFT RIGHT DOWN DOWNDOWN RIGHT LEFT DOWN LEFT understanding of Neuro sci. Psychology DOWN DOWN DOWN human • Subjects change the gazing direction when talk with the sophisticated android. Integration of • Subjects unconsciously feel social relationships with the science and engineering android. Constructive understanding Bottleneck of autonomous robots Understanding of humans as a total system •Previous approaches in brain science Reasoning Development of Humanlike appearance Memory human brain models Humanlike behavior Motor control Analysis of ? & brain computers ? Humanlike perception human brain functions Development of Humanlike conversation? Partial observations Humanlike machine •New approaches in robotics and cognitive & social science • People expect that humanlike robots can Verification of the human likeness talk. in inter-personal and social situations • However, it is a very hard problem… Development of Development of hypotheses humanoids and androids on humanlike systems Constructive approach Systematic understanding of humans An approach for solving the bottleneck of autonomous robots Geminoid Tele-interaction by using semi-autonomous androids and humanoids Humanlike Appearance Android Humanlike behaviors Humanlike perception Operator Internet Humanlike conversation • Long term interaction and conversation • More precise cognitive/psychological tests by using own androids • Exist different places simultaneously 3
  4. 4. Making of the geminoid Definition of geminoid Humanoid = Humanlike robot Android = Robot that has human appearance and behavior Geminoid = Tele-operated android of an existent person Skull MRI image Humanoid Android Geminoid Tele-operation system Developed geminoid through the Internet Internet Difference between my observation and other people’s observations • When I saw the static geminoid, it was like a mirror. However, when it • Motion capture system for measuring the rip movements naturally moved, I could not recognize • Behavior selection by using GUI (6 behaviors) it as my movements. Touch by someone Meeting by using the geminoid rnet Inter Adaptation to the different body • While I operate the geminoid, I unconsciously adapt my movements to the geminoid’s movements. Sharing of information through the geminoid Strong entrainment in the conversation • When the visitor touch to the Geminod, I get a feeling • Both of I and the visitors can quickly, less than 5 minutes, to be toughed (demeaning). adapt to the conversation through the geminoid. 4
  5. 5. Geminoid as a father Habituation in the beginning of the second experiment Father Geminoid of the father Tasks Tasks • Look at pictures together • Look at pictures together • Play last and first • Play last and first • Find a card and give it to the geminoid • Find a card and give it to the • Touch to the geminoid geminoid • Comment on the geminoid Geminoid and children - Children can quickly adapt to the gemioid - New issues in android science Mechanical eng. A. I. Development of mechanical Control sys. Android science humans Robotics Hypothesis •Scientific issue: Human likeness Sensor tec. and verification (appearance, movement, perception) •Engineering issues: simple and Analysis and Cognitive sci. iterative communication tasks understanding of Neuro sci. Psychology human New issues (more philosophical) 1st experiment 2nd experiment • Scientific issue: Human presence How does the self-observation matches to the other’s observations? What is ego? What is human presence? What is authority? Entrainment by conversation and Adaptation to the different body.. Is it possible to separate mind and body? Sharing of information through the geminoid Habituation • Engineering issue: tele-presence technologies using geminoids Synergistic Intelligence Necessity of humanlike mechanisms JST ERATO Asada Project with Prof. Asada Humanlike appearance Humanlike behavior Humanlike perception Humanlike conversation Humanlike mechanism? • The robot needs to have more humanlike mechanism for more flexible movement. Operator Android • How to control the Internet complicated system? • Understanding the developmental process of human/robot through interaction with humans • Realizing developmental systems that includes robots and programmers/ caregivers 5
  6. 6. M3: Growing Man-Made-Man Yuragi Project in Osaka University Center of excellence for new-creating innovations with Prof. Yanagida Self-organizing controller Life Information networks Science System Development robust software for sensory Area Area utilizing “Yuragi” motor mapping Study basic Deliberative behavior principles of biological “Yuragi” Nano Abstraction Material Tree Develop Behavior Area bio-inspired Development of hardware high-performance Reaction for “Yuragi” systems utilizing Sensors Actuators Robotics biological “Yuragi” Area •Develop software by imitating the human developmental process. •Develop fundamental mechanisms that generate a class of cognitive phenomena. Development robotics systems using biological “Yuragi” •The robotic mechanism is used as a hypothesis in cognitive studies. Bio-inspired robotics and Cyborg systems Development of linear actuators Attractors and their switching with Prof. Hirata (Osaka Univ.) activity Idea actuators for the robots Action A Action B •No gears and linear Effective voltage [V] 3.5 Sensor •Flexible control of the stiffness Nu mber of turns [Turns ] 525 Action C Action D •Long stroke Resis tance [Ω] 5 •High torque Pneumatic actuator Mass of mover [g] 238 Internal noise External noise Frict ion force [N] 0.93 Environment Vis cous damping coefficient [ p g [N・s /m] ] 1.0 d x = f ( x ) ⋅ activity + η dt System represented Control parameter Halbach array of magnets Noise with attractors (Sensor) •Biological system utilizes noise. •The complicated system is represented with attractors •Select better solutions by controlling activity •From micro to meta levels Conclusion Robotics is science for understanding humans. Robotics is engineering for developing our future life. Various research areas are integrated in robotics. Information and Robot Society 6