research at the cybernetics lab ima/zlw & ifu rwth aachen ... · 08.12.2016 s. jeschke term:...
Post on 22-Sep-2020
1 Views
Preview:
TRANSCRIPT
www.ima-zlw-ifu.rwth-aachen.de
Research at the Cybernetics Lab
IMA/ZLW & IfU –
RWTH Aachen University
December 2015, Aachen
www.ima-zlw-ifu.rwth-aachen.de
IMA/ZLW & IfU
Faculty of Mechanical Engineering
RWTH Aachen University
2
08.12.2016
S. Jeschke
apl.-Prof. Dr. habil. Ingrid Isenhardt
Deputy Head of Institute
Prof. Dr. rer. nat. Sabina Jeschke
Head of InstituteDr. rer. nat.Frank Hees
Vice Deputy Head of Institute
IMAInstitute of Information Management in Mechanical Engineering
Prof. Dr.-Ing. Tobias Meisen
Managing DirectorJunior Professorship
ZLWCenter for Learning and Knowledge Management
Prof. Dr. phil. Anja Richert
Managing Director
Junior Professorship
IfUAssociated Institute forManagement Cybernetics
Dr. rer. nat. René Vossen
Managing Director
Prof. Dr.-Ing. em. Klaus Henning
Senior Advisor
Traffic and Mobility
Dipl.-Inform. Christian Kohlschein(temporary)
Production Technology
Dr.-Ing. Christian Büscher
KnowledgeEngineering
Sebastian Stiehm M. Sc.
Didactics in STEM Fields
Valerie Stehling M. A.
Agile Mana-gement & eHumanities
Dr.-Ing. ChristianTummel
Engineering Cybernetics
Dipl.-Wirt.-Ing. Sebastian Reuter
CognitiveComputing & eHealth
Dipl.-Inform. Christian Kohlschein
Economic and SocialCybernetics
Dr. phil. Kristina Lahl
Administration IT & Media
Technology
PublicRelations
The Multidisciplinary Institute Cluster IMA/ZLW & IfU
… Organisational Chart
Innovation Research &Futurology
Stefan Schröder M. Sc.
3
08.12.2016
S. Jeschke
ca. 65 scientists
The Multidisciplinary Institute Cluster IMA/ZLW & IfU
… Data and Facts
… at the Faculty of Mechanical Engineering of the RWTH Aachen University
! Cross-disciplinary teams work together on interdisciplinary projects
Team = ca. 215 employees
50 % from Engineering and Natural Sciences50 % from Humanities and Economic Sciences
Gender-Mix: ca. 50:50
ca. 20 employees for technical service and administration
ca. 130 student workers
Culture-Mix: ca. 90:10 (heavily under construction)
ca. 85 % third-party funding
Financial structure in a nutshell
ca. 7 million € annual turnover
4
08.12.2016
S. Jeschke
Term: „governance“, to navigate
Born around 1940
1948: “Cybernetics or control and communication in the Animal and in the machine” (Norbert Wiener)
until 1953: Macy-Conferences
FeedbackloopCircular explanations for systems behavior, self-
regulation(Forrester, Ashby)
AutopoiesisSystem capacity to maintain
and stabilize itself (Maturana, Varela)
DecentralizationDecentralized navigation,
bottom up processes(Stafford Beer)
Central terms in the field of intelligent distributed systems
The central elements of Cybernetics
EmergenceSpontaneous new properties,
swarm behavior (Wolfram, Gell-Mann)
Complex Systems
Multi-component systems in complex interactions
(Stafford Beer)
5
08.12.2016
S. Jeschke
And how do these systems work?
Artificial Intelligence: from GOFAI to Connectivism, about 1980---
Top-Down / symbolic AI
Bottom-Up / subsymbolic AI
Connectivisminteraction as basis
of intelligence
[Szt
ipan
ovi
ts,
Van
der
bilt
, 19
97
]
[Lin
, 20
10
]
Knowledge on demand /knowledgeaquisition
GOFAIGood old fashionedArtificial Intelligence; based on high-level "symbolic“ knowledgerepresentations
Knowledge storage/
knowledgeretrieval
Two competing movements? --- From top-down to button-up design
6
08.12.2016
S. Jeschke
… leading to the 4th industrial (r)evolution...
Breakthroughs - A new era of artificial intelligence
Communication technologybandwidth and computational power
Embedded systemsminiaturization
Semantic technologiesinformation integration
Google Car2012
Systems of “human-like” complexity
Artificial intelligencebehavior and decision support
1st dimension: complexity level
Watson 2011
7
08.12.2016
S. Jeschke
… leading to the 4th industrial (r)evolution...
Breakthroughs - Everybody and everything is networked
Team Robotics
Swarm Robotics
Smart Grid
Smart Factory
Car2Infra-structure
2n
dd
ime
nsi
on
: n
etw
ork
leve
l
Communication technologybandwidth and computational power
Embedded systemsminiaturization
Semantic technologiesinformation integration
Artificial intelligencebehavior and decision support
8
08.12.2016
S. Jeschke
Communication technologybandwidth and computational power
Embedded systemsminiaturization
Semantic technologiesinformation integration
… towards a networked world
And how do these systems work?
Power revolutionCentralized electric power infrastructure; mass production by division of labor
1st industrial revolutionMechanical production systematically using the power of water and steam
today
Digital revolutionDigital computing and communication technology, enhancing systems’ intelligence
Information revolutionEverybody and everything is networked – networked information as a “huge brain”
around 1750 around 1900 around 1970
Towards intelligent and (partly-)autonomous systems AND systems of systems
?? Steering -Controlling ??
9
08.12.2016
S. Jeschke
Power RevolutionCentralized electric power infrastructure; mass production by division of labor
1st Industrial RevolutionMechanical production systematically using the power of water and steam
Today
Digital RevolutionDigital computing and communication technology, enhancing systems’ intelligence
Information RevolutionEverybody and everything is networked – networked information as a “huge brain”
Characteristics of Industrial Revolutions:
The vendor change
Around 1750 Around 1900 Around 1970
Latest version of Google’s self driving car (Huffington Post, 28.5.2014)
Google: First autonomic car with street license, 2012
Ford 021C concept car 2012, designed by Newson now at Apple (1999)
Apple Inc.
Innovations in 4.0
The vendor change around „cars“
Tesla X 2015, other Teslassince 2006; Forbes: “most innovative enterprise”
Sony announced autonomous car in 2015, based on their experience in visual sensors
Car specialists? – No. Connectivity & data
specialists. Energy & sensor
specialists.
10
08.12.2016
S. Jeschke
Power RevolutionCentralized electric power infrastructure; mass production by division of labor
1st Industrial RevolutionMechanical production systematically using the power of water and steam
Today
Digital RevolutionDigital computing and communication technology, enhancing systems’ intelligence
Information RevolutionEverybody and everything is networked – networked information as a “huge brain”
Characteristics of Industrial Revolutions:
The vendor change
Around 1750 Around 1900 Around 1970
Latest version of Google’s self driving car (Huffington Post, 28.5.2014)
Google: First autonomic car with street license, 2012
Ford 021C concept car 2012, designed by Newson now at Apple (1999)
Apple Inc.
Innovations in 4.0
The vendor change around „cars“
Tesla X 2015, other Teslassince 2006; Forbes: “most innovative enterprise”
An autonomous car is more like a computer on wheels than a car which includes one or many computers.
Sony announced autonomous car in 2015, based on their experience in visual sensors
11
08.12.2016
S. Jeschke
The creative artificial mind
Industry 4.0 does not only change the “routine” jobs
The typical assumption…
… that job changes in 4.0 are mainly addressing blue collar jobs and/or routine jobs does not hold true.
White collar jobs
… are under massive change due to the enhancement in
AI, here the impact often hits “middle class jobs”
Social robots
… will become capable of taking over even complex
tasks with personal presence as in health or home care
From „blue collar – low qualified“ to „white collar – middle class“...
but probably, this is just a transition phenomenon
High qualified jobs
… as e.g. health professionals face already the taking over
through AI in certain fields by Watson, Google Flu, etc.
Decentralized platforms
… with automated consensus models (e.g. blockchain) take over complex administrative
tasks e.g. in judiciaries
Virtual and augmented environments
… allowing for new international players, even in tasks requiring humans
and presence
Autonomous systems
… as autonomous cars and more envanced production
technology will change the blue collar – low qualified as well
12
08.12.2016
S. Jeschke
From the history of autonomous vehicles
2009: Truck robot platoons – distributed intelligence
The KONVOI project (several institutes from RWTH & industry partners)
2005-2009 automated / partly autonomous
transportation e.g. by electronically coupling trucks to convoys
several successful tests with trucks: Chauffeur, KONVOI, SARTRE (EU), Energy-ITS (Japan), …
expected improvements:beyond safety, reduction of fuel consumption and gained road space
!
Adv. driver assistance system for trucks short distances between vehicles of
approx. 10m at a velocity of 80 km/h Energy-ITS: 4m ! (2013)
KONVOI: Car2infrastrcuture components! Model of multi agent systems
Connectivity…
13
08.12.2016
S. Jeschke
Organization forms on demand – individualized by client – initialized by product
!Product agitates as “super-agent”: Plans production and transportation steps Requests services from agents Negotiates with other products for agent-resources
Heterogeneous player modeled as multi agent concept Models from biology and social sciences Based on autopoiesis & embodiment theory
Transport unit
Production unit
Virtual service providerFa
bri
cati
on
Ou
tsid
e w
orl
d
© D
aniel Ew
ert 20
13
Changes already „under construction“
With decentralized models towards lot size 1
Konvoi 2005-2009, RWTH with partners
(partly) autonomous driving via convoys
14
08.12.2016
S. Jeschke
Advantage of decentralized control structures
Intralogistics goes mobile: The Festo Logistics League
Competitions robocup:
2012: 0 points in World Cup
2013: 4th in World Cup
2014: Winner of the GermanOpen
2014: Winner of the World Cup
2015: Winner of the World Cup
2016: Winner of the World Cup
Critical factors for success: Totally decentralized No ”hard coded components“ Strong cooperation Re-planning during tasks
Mobile transportation robots from flexible routing
!
Competencies: localization & navigation computer vision adaptive planning multi agent strategies sensory & hardware
15
08.12.2016
S. Jeschke
Lend the robots a face
Into Service Robotics: The next step – the “Oscars”
Transform mobile robotic experiences into the field of service robotics
!
1. Investigating “new” human machine Interfaces and interaction schemes Simple, intuitive Schematic eyes following you “natural eyes behavior”: randomly
looking around, showing interest by blinking, looking bored, …
!Performing service robot tasks Distribute brochures and serving drinks Path planning, room exploration, …
!
2. Investigating the “Uncanny Valley”: when features look almost, but not exactly, like natural beings, it causes a response of revulsion among the observers (Mori 1970)
3. Investigating diversity specific reactions (gender, age, culture) to artificial systems and in particular robots
16
08.12.2016
S. Jeschke
Data-driven learning – supervised
Supervised learning in high-pressure die casting
? Can we predict the result of a HPDC (high-pressure die casting) process –by using historical data? – YES, WE CAN! 2015
HPDC process measurements
Historic dataProcess and quality data
Prediction modelModelling and
training
Visualization of predictionInline and web-based
(result NIO|IO with reason)
IOIO
NIO (Outbreak)
NIO (Blowhole)
NIO (Cold shot)
Setting: The die casting equipmentin the research wing was separatedfrom the quality check. Thus, the forms were checked with a delay in time and a considerable spacialshift. Under these conditions, „reproducibility“ of the resultscould not be reached...
Not the original
… in cooperation with
17
08.12.2016
S. Jeschke
Data-driven learning – supervised
Supervised learning in individualized production (1)
? Can we use supervised learning to improve very flexible production processes towards lot size 1? – YES, WE CAN!
Injection Moulding / synthetics (at IKV):
For lot size 1, real experiments are expensive Current simulations in the field help to optimize
the results but do not feedback with the process At IKV, a large amount of training data is available High quality process simulations available
Automating the moulding process alongparameters coming from geometry, weight, etc.
Hot Rolling /metals (at ibf):
Coils or slabs depend highly on the costumer So far, the rolling schedule is driven by
experience rather than by automated design Process simulations of mill trains are “physically
incomplete”, real experiments are expensive
Automating the rolling schedule alongparameters as the number of millings, pressureapplied, and temperature
! The BRAIN PROJECT - by the cluster of excellence (since 9/2016)
18
08.12.2016
S. Jeschke
Sulfur (S) > 0.04% and heat treatment fragile structure
Phosphorus (P) > 0.04% reduced plasticity
Chrome (Cr) > 16%, Molybdenum (Mo) > 13%, Nickel (Ni) > 56% no findings
…
Data-driven learning – unsupervised
Unsupervised learning in a somewhat “corrupt” database
!Finding hidden relations in our data, we were not aware of, e.g., understanding failures or bad quality of products and processes 2015
… in cooperation with
Searching for hidden relations in data by
subgroup mining
Unsupervised cleaning
Setting: The company was facing a certain muddle in theirexpert knowledge base. If all the information providedwas to be true – then certainphenomenons would not occur. But did.
The suspicion was thatsome information might beoutdated by now, but the specialists still believe them...
But – which one?
Data about chemical compositions
19
08.12.2016
S. Jeschke
Data-driven learning – unsupervised
Unsupervised learning for laser cutting processes
?Can we use unsupervised learning to identifying a group of desired process results in a highly complex process - YES, WE CAN!
! Part of B1 - by the cluster of excellence (since about 1/2016)
Sheet Metal Cutting (at NLD):
challenge: reaching a maximum assurance of the cut quality reliable process simulations are available (QuCut, developed at NLD) thus, a large amount of training data can be generated
Automating the cutting process along parameters coming from
five optics design parameters (beam quality, astigmatism, focal position, beam radius in x and y directions (elliptical laser))
and eight process criteria (roughness values at different depths)
[all pictures/movies:NLD, W. Schulz, 2016]
20
08.12.2016
S. Jeschke
Data-driven learning – unsupervised
… using k-means clustering
Results of cluster analysis:• Identification of desired simulation runs (i.e. blue cluster)• Further analytics reveals process parameters that lead to desired results
multidimensional visualization
K=5 clusters... ... transferred into parallel coordinates(reduced to the 8-dim. roughness space)
k-means clustering groups n observations into k clusters. Using k-means clustering on simulation data allows to… analyze multidimensional relationships between process criteria discover hidden structures in experimental results gain knowledge about the structure in the data
2nd paper currentlyunder review at wgp
?Can we use unsupervised learning to identifying a group of desired process results in a highly complex process - YES, WE CAN!
! Part of B1 - by the cluster of excellence (since about 1/2016)
21
08.12.2016
S. Jeschke
Deep learning combined with reinforcement learning
Deep learning to deal with variations
? Can we make a robot really understand its task and perform movements that solve this task adapting to dynamic and variable circumstances of the real world?
… in cooperation with
Setting: An industrial robot with a veryelastic joint is to grab objects from a distinct location a and pitch them into a hole at location b. Here, the shapes of the objects possess a certain variation – like „real bananas“. Traditional programming is inflexible –slightest variations leads to failure of the task.Can we make the robot learnto grab and put different kindof objects?
An industrial robot... training for „crooked cylinders in the box“
Pretraining to reduce the amount of experience needed to train
visuomotor policies Enhance security for robot and environment as it
limits the motion to a certain “frame”
Neural network of 7 (4) layers (3 convolutional layers for vision, skipped here) 1 expected position layer that converts pixel-wise
features to feature points 3 fully connected layers to produce the torques.
Main training process
Model predictiveregulators
Optimization of the dynamics model
Max. of thereward
Generate trainingdata of the model
Neuronal Network
Dynamics model
Pretrainingprocess
Source: Levine, S., Wagener, N., & Abbeel, P. (2015, May). Learning contact-rich manipulation skills with guided policy search. In 2015 IEEE international conference on robotics and automation (ICRA) (pp. 156-163). IEEE.
Not the original
22
08.12.2016
S. Jeschke
Deep learning combined with reinforcement learning
Deep learning to replace teach-in programming
? Can we use deep learning to improve flexible motion patters for robots –all towards lot size 1? – YES, WE CAN!
! The CENSE PROJECT – by the cluster of excellence (since 5/2016)
Machine tools
Robot playing the „wire loop game“ (at IMA & WZL):
challenge: guiding a metal loop along a twisted wire (geometry unknown) without touching the loop to the wire
reliable simulations are available, which allows for first training in VR before experimenting in the real world
Automating the motion path along different wires
Getting rid of the teach-in programming as it is not flexible enough for indivudualized production
Transfer to the process of production and assembly Planning of welding lanes Applying of glue beads Assembly of limp components
1st paper currentlyunder review at CIRP
23
08.12.2016
S. Jeschke
Deep learning combined with reinforcement learning
Deep learning to replace teach-in programming
? Can we use deep learning to improve flexible motion patters for robots –all towards lot size 1? – YES, WE CAN!
! The CENSE PROJECT – by the cluster of excellence (since 5/2016)
Machine tools
Environment model
Environmental state
Q-Learning (special caseof R-Learning)
Deep neural network
Communication via web service connection
𝒙𝟏𝒙𝟐𝒙𝟑𝒙𝟒...𝒙𝒏
Feature vectors
𝒙𝟏𝒙𝟐𝒙𝟑𝒙𝟒...𝒙𝒏
𝒙𝟏𝒙𝟐𝒙𝟑𝒙𝟒...𝒙𝒏
𝒙𝟏𝒙𝟐𝒙𝟑𝒙𝟒...𝒙𝒏
Simulation environment
Next step: Improving topology towards CNN(convolutional neural network)
24
08.12.2016
S. Jeschke
The creative artificial mind
Where the Story Goes: AlphaGo
!
Go originated in China more than 2,500 years ago. Confucius wrote about it. As simple as the rules are, Go is a game of profound complexity. This complexity is what makes Go hard for computers to play, and an irresistible challenge to AI researchers. [adapted from Hassabis, 2016]
Bringing it all together!
The problem: 2.57×10210 possible positions – that is more than the number of atoms in the universe, and more than a googol times (10100) larger than chess.
Training set30 million moves recorded fromgames played by humans experts
Creating deep neural networks12 network layers with millions ofneuron-like connections
Predicting the human move(57% of time)
Dat
a-d
rive
n le
arn
ing
Re
info
rce
me
nt
lear
nin
g
Learning non-human strategiesAlphaGo designed by Google DeepMind, played against itself in thousands of games and evolved its neural networks; Monte Carlo tree search
! Achieving one of the grand challenges of AI
March 2016:Beating Lee Se-dol (World Champion)AlphaGo won 4 games to 1.(5 years before time)
[Has
sab
is, 2
01
6]
25
08.12.2016
S. Jeschke
The creative artificial mind
From discrete to continuous variance
From the paradigm „Individualization by building blocks“... Consumers chose from predefined options Processes are separately planned and realised for each option
2017
... Towards the paradigm „Individualization by continuous models“ Consumers taylor their individual (non-prodefined) product Processes are carried out in real-time, within a model
2030
Customer specifying car toindividual needs; at home; incl. „continuous features“
+ 20 cm, drivers seat
area
CAD & Design CAM
Manufacturing needed tools
part manufacturingAssembly
AI drivenfeedback loops
Delivery ofindividualized car
26
08.12.2016
S. Jeschke
Head of Institute:Univ.-Prof. Dr. rer. nat. Sabina JeschkeTel.: +49 241-80-91110sabina.jeschke@ima-zlw-ifu.rwth-aachen.de
www.ima-zlw-ifu.rwth-aachen.de
Deputy Heads of Institute:
apl.-Prof. Dr. habil. Ingrid IsenhardtTel.: +49 241-80-91112ingrid.isenhardt@ima-zlw-ifu.rwth-aachen.de
Dr. rer. nat. Frank HeesTel.: +49 241-80-91113frank.hees@ima-zlw-ifu.rwth-aachen.de
27
08.12.2016
S. Jeschke
1968 Born in Kungälv/Sweden
1991 – 1997 Studies of Physics, Mathematics, Computer Sciences, TU Berlin1994 NASA Ames Research Center, Moffett Field, CA/USA10/1994 Fellowship „Studienstiftung des Deutschen Volkes“ 1997 Diploma Physics
1997 – 2000 Research Fellow , TU Berlin, Institute for Mathematics2000 – 2001 Lecturer, Georgia Institute of Technology, GA/USA2001 – 2004 Project leadership, TU Berlin, Institute for Mathematics04/2004 Ph.D. (Dr. rer. nat.), TU Berlin, in the field of Computer Sciences2004 Set-up and leadership of the Multimedia-Center at the TU Berlin
2005 – 2007 Juniorprofessor „New Media in Mathematics & Sciences“ & Director of the Multimedia-center MuLF, TU Berlin
2007 – 2009 Univ.-Professor, Institute for IT Service Technologies (IITS) & Director of the Computer Center (RUS), Department of Electrical Engineering, University of Stuttgart
since 06/2009 Univ.-Professor, Head of the Cybernetics Lab IMA/ZLW & IfU, Department of Mechanical Engineering, RWTH Aachen University
2011 – 2016 Vice Dean of the Department of Mechanical Engineering, RWTH Aachen University
since 03/2012 Chairwoman VDI Aachen
since 05/2015 Supervisory Board of Körber AG, Hamburg
Prof. Dr. rer. nat. Sabina Jeschke
29
08.12.2016
S. Jeschke
Data-driven learning – supervised
… optimizing the classifiers
? Can we predict the result of a HPDC (high-pressure die casting) process –by using historical data? – YES, WE CAN! 2015
… in cooperation with
! Classification – “One classifier to rule them all?”Different classification models can be used to categorize the states of your system.
𝒙𝟏𝒙𝟐𝒙𝟑𝒙𝟒...𝒙𝒏
Observation(feature vector)
𝒚𝟏𝒚𝟐...𝒚𝒎
Probabilityof category
Classifier: kNN classifier decision trees random forest tree support vector machines naïve Bayes classifier …
73%
85%
98%Setting: The die casting equipmentin the research wing was separatedfrom the quality check. Thus, the forms were checked with a delay in time and a considerable spacialshift. Under these conditions, „reproducibility“ of the resultscould not be reached...
… in cooperation with
30
08.12.2016
S. Jeschke
Cyber-physical systems
… and related terms: Internet of Things & Industry 4.0
Cyber-physical systems OR Internet of things ?
• large-scale distributed computing systems of systems• Computation and “intelligence” is
not decoupled from environment
• Internet as large-scale network • embedded systems
(= intelligent components)
Shared
Vision
Core Technology
• Internet of Thingsdriven from computer sciences, Internet technologiesdriven by EC
• Cyber-physical systemdriven from engineering aspects driven by the NSF
• Internet of Thingsfocusing on openness and on the network - virtuality
• Cyber-physical systemfocusing on the physical system behind, often a closed-loop system
Distinct
Scientific Community
Philosophy, focus
For all practical purposes:• Today: more or less synonym• Industry 4.0 as a special field of application
31
08.12.2016
S. Jeschke
The creative artificial mind
Creativity as basis for innovation
!“Creativity is a phenomenon whereby something new … is formed. The created item may be intangible (such as an idea, a scientific theory, a musical composition or a joke) or a physical object (such as an invention, a literary work or a painting).” [adapted from Wikipedia, last visited 5/3/2016]
„In 1917, Sigmund Freud offered a theory of the three mortifications of man:
1. the cosmological (Kopernikus, man isn’t the center of the universe),
2. the biological (Darwin, man descends from the ape)
3. and the psychological (Freud himself, man isn’t master of his own psyche).
In line with the rapid development of Artificial Intelligence, the addition of another mortification suggests itself:
4. the intellectual: there are more intelligent entities than man.“
[Dr. phil. Kristina Lahl, RWTH, Cybernetics Lab]
Rather, it appears to be a mixture between “wishful thinking” and a
lack of definition.
The frequently uttered assumption that AI can’t be creative does not hold true.
32
08.12.2016
S. Jeschke
neuromorphiccomputing
machinelearning
The “second way”
About the connection between birds and AI
“You don’t need to be a bird to fly.”
!
In traditional machine learning, “general purpose computers” are considered.
These computers have no strong similarities of biological brain structures.
Even if they work less effective than biological brains (so far), they have a enormous energy consumption.
!
In the “brain projects”, specialized computer architectures are developed, driven by biological paradigms.
These architectures are more efficient for certain tasks, but do not follow the “general purpose idea” any longer.
Hardware and software become strongly coupled. Thus, experimental changes become more complicated.
… however, it‘s ok to be a bird to fly!
33
08.12.2016
S. Jeschke
Limits of general purpose computers and machine learning
The limitations of the Moore’s law
Gordon Moore (born 1929) co-founder and Chairman Emeritus
of Intel Corporation before: director of R&D at Fairchild
Semiconductor education: Berkeley, Caltech, Hopkins
Moore's law(from a speech given by Moore, in 1965)
… the prediction that the number of transistors that can be placed on the same volume and for constant costs will double every 18 months!
(above: the current definition; the original used a doubling speed of one year)
5 – the magicboundary…
34
08.12.2016
S. Jeschke
Limits of general purpose computers and machine learning
Von Neumann architecture
John von Neumann (1903 – 1957) a Hungarian-American mathematician,
physicist (quantum mechanics), inventor, computer scientist, …
Manhattan project ENIAC project: “Great brain”, first pure
electronic computer, 1946 turing complete
!
Turing completeness:
colloquial: “a general-purpose computer can do what a turing machine can do”
And by Church’s lemma: “a turing machine can compute ‘every algorithm’”
thus, Turing completeness means: can compute everything as long as the algorithm is known
Von Neumann architecture is turing -complete
The von Neumann bottleneck:
The shared single bus between the program memory and data memory
limits the data transfer rate between CPU and memory
Resulting in an intellectual bottleneck: “only one thing at a time thinking”
Memory wall: CPU speed rises faster than the speed of transfer and in the memory
35
08.12.2016
S. Jeschke
Towards new approaches in data sciences
What is neuromorphic computing?
“Neuromorphic engineering, also known as neuromorphic computing, … describing the use of very-large-scale integration (VLSI) systems containing electronic analog circuits to mimic neuro-biological architectures present in the nervous system.
Neuromorphic engineering is an interdisciplinary subject … to design artificial neural systems, such as vision systems, head-eye systems, auditory processors, and autonomous robots, whose physical architecture and design principles are based on those of biological nervous systems.
36
08.12.2016
S. Jeschke
Towards new approaches in data sciences
The “Brain projects”
! Brain Initiative
White House BRAIN Initiative (Brain Research through Advancing Innovative Neurotechnologies)
Established: 2013 by Obama Timeline: about 10 years
Directors: Cornelia Bargmann and William Newsome
Web: http://www.braininitiative.nih.gov/
total costs: private-public mix, about 100 mio $ in 2014, …
!
Human Brain Project (HBP)
EC FET Flagship
Established : 2013 Timeline: about 10 years
Director: Henry Markram Coordinator: EPFL Web: www.humanbrainproject.eu Partners: 100 all over Europe, in Germany:
Heidelberg, Jülich, …
total costs: 1.19 billion €, about half of it provided by the EC
? Worldwide, already a strong competition has been started.The biggest research activities in this field are … :
37
08.12.2016
S. Jeschke
Towards new approaches in data sciences
Examples for neuromorphic chips
Intel, sensor image processing architecture
SpiNNaker, component of the Human Brain Project
!
SpiNNaker: Spiking Neural Network Architecture
Spiking: inculdes time and temporal coding novel computer architecture goal: to use 1 mio. ARM processors (currently
0.5 mio) in a massively parallel computing platform based on spiking neural networks
“One Million Chips Mimic One Percent Of The Brain”
By Steve Furber, Univ. of Manchester, one of the world’s best microprocessor designers
!
Intel Reveals Spin-based Neuromorphic Chip Design with up to 300 times lower energy computation
Involves the combined use of spintronics and memristors (memory resistors, not constant but depend on the history of current). In a cross-bar switch lattice, lateral spin valves act as neurons, and memristors act as synapses.
[a paper published by Intel in June 2012, Sharad et.al, 2012: arXiv:1206.3227]
38
08.12.2016
S. Jeschke
Towards new approaches in data sciences
Timeline – results so far and the expectations in the future
? After Moore’s law: What might be the replacement?A pipeline of new technologies to prolong Moore’s magic
“For some time, making transistors smaller has no longer been making them more energy-efficient; as a result, the operating speed of high-end chips has been on a plateau since the mid-2000s.”
[economist, 2016]
!
The short history of transistors
• 2004 strained silicon• 2007 metal oxides used to beat the
effects of tunneling• 2012 finFET transistors• 2020 “gate-all-around“ transistors
Beyond changes in the design of transistors, more exotic solutions may be needed? For example materials beyond silicon. But in the end, we just “stave off the need for something radical”.
[Greg Yeric, ARM transistor designer, 2016]
39
08.12.2016
S. Jeschke
Is AI at its end because Moore is not holding much longer?
NO. New technologies will take over, however there may be a gap in time…
NO, because additionally, the future may lay in more and more powerful algorithms instead of pure computational speed and power.
Specialized chipsneuromorphic computing
New materials nanotubes, cadmiumtellurid,
graphene, molybdenite, …
Quantum computingmake direct use of quantum-
mechanical phenomenaNew geometries3D architectures, …
Towards new approaches in data sciences
In a nutshell, the alternatives to the beaten paths
40
08.12.2016
S. Jeschke
Summary
The “new AI”: Cognitive Computing
“Cognitive computing (CC) makes a new class of problems computable. It addresses complex situations that are characterized by ambiguity and uncertainty; in other words it handles human kinds of problems. …To do this, systems often need to weigh conflicting evidence and suggest an answer that is “best” rather than “right”.Cognitive computing systems make context computable.”
“Cognitive computing systems [are] a category of technologies that uses natural language processing and machine learning to enable people and machines to interact more naturally […]. These systems will learn and interact to provide expert assistance to scientists, engineers, lawyers, and other professionals in a fraction of the time it now takes.”
“Cognitive computing is the simulation of human thought processes in a computerized model…. involves self-learning systems that use data mining, pattern recognition and natural language processing to mimic the way the human brain works.”
41
08.12.2016
S. Jeschke
New forms of human machine interaction
About the role of emotion in human-maschine-interaction
!However, it took a while before emotions wereconsidered important in computer science.
Plato (ca. 400 BC)“Human behavior flows from three main sources: desire, emotion, and knowledge.”
Rosalind Picard (since 1997; MIT)“Computers that will interact naturally and intelligently with humans need the ability to at least recognize and express affect.”
Picard coined the term“affectice computing”
KISMET - MIT (1990-2000; Cynthia Breazeal)
Analysis and simulation of human-like emotions
Research on interaction between robotsand humans
Part of the “organic development”
42
08.12.2016
S. Jeschke
New forms of human machine interaction
Automatic Emotion Recognition: Project IMOTION (1)
?Did you ever yell at your GPS, e.g. when it told you the equivalent of “drive through that river ahead!”?
Well, WE do constantly. And so the story went on like this:
! Motivation and goal: Transform the GPS into an intelligent co-driver, i.e. get it to adapt to your emotions!
Solution: A machine-learning based system architecture. Did it work? Often! Was it fun? Hell, yeah!
Ingredients in a nutshell
Primary source of emotion: driver speech
Machine-learning algorithm: Support Vector Machine (SVM)
Training database: talk show based corpus (lots of yelling)
Test-bed: a driving simulator
Test persons: old and young
43
08.12.2016
S. Jeschke
Changes already „under construction“
Towards human-robot cooperation: hybrid teams
Towards hybrid teams and in-the-box production
© F.Welter Aachen © F.Welter Aachen
Audis collaborative robots in Ingolstadt, the “Cobots” pick up components and pass them to workers (02/2015)
New intelligence models: New AI for “context understanding”
New “body concepts” for robots New types of “sensible” robots, mainly “lightweight”
Real-time capability: New fast sensors allows avoiding accidents in close cooperation
PhD Ying Wang, RRWTH, IMA/ZLW & IfU, 2016
44
08.12.2016
S. Jeschke
Some even more “out of space” concepts
The new driver
!
„My colleague the robot…“
Again more: In a few years, automated driving might outcompete human drivers.Security issues, the demographic change, and the decreasing attractiveness of the job may add to a fast change.
DHL
DaimlerGoogle
www.cargocap.de
www.spektrum.de
Rolls Roycerail-bound caps
45
08.12.2016
S. Jeschke
Some even more “out of space” concepts
The third dimension
The megacities of the future
At a certain point, due to purely mathematical reasons, extended 3-dimensional building structures can not longer be served by purely 2-dimensional (street) networks.
Source : National Geographic Magazine
!
“In 2030, 70% of all humans will live in cities. Already then, about 10% will live in megacities (i.e., more than 10 Mio people). Escalating…”
[freestyle translation, source pwc studies]
Below ground
Above ground
Source : Cargo sous terrain, CH
Third dimension
46
08.12.2016
S. Jeschke
“plastics instead of parcels?” -UPS moving from logistics to3D printing (tests since 2013)
Some even more “out of space” concepts
The new construction
!
„Digital warehouses are replacing physical spare parts storages“[freestyle translation, source Logistik magazine]
„3D printing is on its way to leave the somewhat ‘restricted’ areas of spare part business, tool making etc. and is about to become a serious challenger for all traditional manufacturing models“.
[source Prof. Erman Tekkaya, TU Dortmund]
Harbor Rotterdam – 3D printer farm for metal printing (after piloting, now roll-out in 2016)
Water carbonatorsreaching high salesfigures
3D printing of house (sourceUniv. of Southern California 2013)
3D print of pasta –Barilla (tests since 2015)
47
08.12.2016
S. Jeschke
Service Innovations
Expectation management – from B2C to B2B and the Generation Y!
The costumer gets powerful.
And: he/she expects services in business (B2B) to work in the same comfortable way as at home (B2C).
And – more again: the digital native is entering the scene. This guy does not even know what a fax machine is used for. Everything outside the internet does not exist!
Business units – such as marketing, sales, customer support – communicate with each other, but also directly and autonomously with the customer.
Marketing Customer Support
Sales
Business Customer
Customer focusedmulti-channelcommunication
Traditional enterpriseCommunication
Enabled byservice-orientedbusiness models
Internal communicationbecomes more efficient
48
08.12.2016
S. Jeschke
!Services become available and experts become obsolete!
As information becomes more and accessible, experts lose information power. This observation is inline with all earlier changes along the information chain, starting with the book printing …
Expert systems losing ground: Experts and service providers lose their unique selling proposition
ExpertKnowledge
Tourist Guide Real Estate
Salesman
Travel Agent
Informed Customer Profound Decisions
Management of services: The new business model
Service InnovationsBusinesses in the customer’s pockets: the „Service Society“
Information and knowledge
transparency
expertise“replaced” by information
Smartphonesingle point of contact for all information
49
08.12.2016
S. Jeschke
CarrierRequest
Delivery Customer Delivery Service End Consumer
Requests and Status Updates
!
Logistics 4.0 or „Logistics as a Service“ (LaaS)
The terminology is based on concepts as Software-as-a-Service (SaaS), Platform-as-a-Service (PaaS) …. up to Anything-as-a-Service (XaaS).It is sometimes referred to as "on-demand XY”, without hosting or owning the necessary infrastructures and tools.
The philosophy behind it is: “Just do it – I don’t care how!”
ManagingPlatforms
MobileApps
Service Innovations
… dealing to „Logistics as a Service“ (LaaS)
B2B relations still lacking generic services like in B2C relations
50
08.12.2016
S. Jeschke
Service Innovations
From production to services: The shift towards a sharing economy
! Uncertain future of patents… IPR?
Sharing economy: “A common premise is that when information about goods is shared (typically via an online marketplace), the value of those goods may increase for the business, for individuals, for the community and for society in general.”
[Wikipedia, 2015]
„The sharing economy endangers traditionalbusiness models. E.g. carsharing could choke offthe demand of new cars…“
[http://www.haufe.de/ 2015]
!
“…already today, about 25% of the Germans can becounted as `social-innovative co-consumer`“
[Heinrichs, Leuphana, 2015: Auf dem Weg in eine
neue Konsumkultur?]
top related