Large-scale projects to build Large-scale projects to build artificial brains: review. artificial brains: review.

Włodzisław Duch (Google: Duch)Włodzisław Duch (Google: Duch)

Department of Informatics, Department of Informatics, Nicolaus Copernicus University, Nicolaus Copernicus University,

Torun, PolandTorun, Poland

School of Computer Engineering, School of Computer Engineering, Nanyang Nanyang TechnologicalTechnological University (NTU), University (NTU),

SingaporeSingaporeBuilding Artificial Brain – workshop after ICANN 2005, Sept 15,

2005

Plan• Motivation: are we ready for brain simulation? Motivation: are we ready for brain simulation? • Some failed attempts.Some failed attempts.• Special hardware? Special hardware? • Nomad/Darwin robots, Gerald EdelmanNomad/Darwin robots, Gerald Edelman• Blue Brain – Henry Markram, Lausanne/IBMBlue Brain – Henry Markram, Lausanne/IBM• CCortex, Artificial Development.CCortex, Artificial Development.• The Ersatz Brain Project, James AndersonThe Ersatz Brain Project, James Anderson• Ai – developing brains? Ai – developing brains? • Conscious machines: Pentti Haikonen (Nokia) & others.Conscious machines: Pentti Haikonen (Nokia) & others.• Bayesian confidence propagating network: Lansner Bayesian confidence propagating network: Lansner • Artificial Mind System – Testuya HoyaArtificial Mind System – Testuya Hoya• NTU projects in artificial minds NTU projects in artificial minds • Related EU projects and initiativesRelated EU projects and initiatives• Related: consciousness is not that hard; how to get mind out of Related: consciousness is not that hard; how to get mind out of

brain?brain?

Motivation: developments in computingMotivation: developments in computing

Naive estimation of the brain power: Naive estimation of the brain power: BP = 100 Hz x 10BP = 100 Hz x 101414 synapses = 10 synapses = 101616 binop/s. binop/s.

Power for abstract thinking is probably much lower. Power for abstract thinking is probably much lower.

Kasparov lost in 1997 with Deep Blue machine that searched 200M Kasparov lost in 1997 with Deep Blue machine that searched 200M nodes/sec, less than 10nodes/sec, less than 101212 binop/s, on 32-processor IBM SP + 512 binop/s, on 32-processor IBM SP + 512 specialized chess processors. This gives about 0.01% of BP. specialized chess processors. This gives about 0.01% of BP.

Kramnik (2002) reached a draw with 8-processor Windows XP Kramnik (2002) reached a draw with 8-processor Windows XP machinemachine running commercial version of Deep Fritz program. running commercial version of Deep Fritz program.

Supercomputer speeds have just reached > 100 Tflops, or a few Supercomputer speeds have just reached > 100 Tflops, or a few Petaops/sec, comparable with brain power, Grid computing arrived, but Petaops/sec, comparable with brain power, Grid computing arrived, but computers are far from brain’s complexity and processing style. computers are far from brain’s complexity and processing style.

In the near future 1000$ PC will have brain power.In the near future 1000$ PC will have brain power.

CComputingomputing/inteligence /inteligence

Computing costsComputing costs

Motivation: neuroscienceMotivation: neuroscience

From the “Blue Brain” project: From the “Blue Brain” project:

Scientists have been accumulating knowledge on the structure and Scientists have been accumulating knowledge on the structure and function of the brain for the past 100 years. It is now time to start function of the brain for the past 100 years. It is now time to start gathering this data together in a unified model and putting it to the test gathering this data together in a unified model and putting it to the test in simulations. We still need to learn a lot about the brain before we in simulations. We still need to learn a lot about the brain before we understand it's inner workings, but building this model should help understand it's inner workings, but building this model should help organize and accelerate this quest.organize and accelerate this quest.

The data obtained on the microstructure and function of the NCC has The data obtained on the microstructure and function of the NCC has now reached a critical level of detail that makes it possible to begin a now reached a critical level of detail that makes it possible to begin a systematic reconstruction of the NCC. The numbers and types of systematic reconstruction of the NCC. The numbers and types of neurons have basically been defined, who connects to whom and how neurons have basically been defined, who connects to whom and how often, has been worked out, and the way that most of the neurons often, has been worked out, and the way that most of the neurons function as well as the way that the neurons communicate and learn function as well as the way that the neurons communicate and learn has been extensively studied. has been extensively studied.

We therefore now have a near complete digital description of the We therefore now have a near complete digital description of the structural and functional rules of the NCC. structural and functional rules of the NCC.

Scheme of the brain ...Scheme of the brain ...

High-level sketch of the brain structures, with connections based on High-level sketch of the brain structures, with connections based on different types of neurotransmiters marked in different colors.different types of neurotransmiters marked in different colors.

Motivation: more science Motivation: more science

• Engineering: to be sure that we understand complex system Engineering: to be sure that we understand complex system we need to build and test them.we need to build and test them.

• Understanding emergent properties of neural systems: how Understanding emergent properties of neural systems: how high-level cognition arises from low-level interactions between high-level cognition arises from low-level interactions between neurons. neurons.

• Removing all but a few areas of the brain will to lead to Removing all but a few areas of the brain will to lead to functional system, therefore even crude simulation that functional system, therefore even crude simulation that includes all major areas can teach us something. includes all major areas can teach us something.

• Build powerful research tool for brain sciences.Build powerful research tool for brain sciences.• So far the only architecture of cognition is SOAR, based on So far the only architecture of cognition is SOAR, based on

the idea of physical symbol processing system, originated by the idea of physical symbol processing system, originated by Newell, Simon & developed over the last 25 years. SOAR and Newell, Simon & developed over the last 25 years. SOAR and ACT-R were very successful in explaining different features of ACT-R were very successful in explaining different features of behavior and used in problem solving although they little to do behavior and used in problem solving although they little to do with brain-like information processing. with brain-like information processing.

Motivation: practicalMotivation: practical

Large computer power allows for building Large computer power allows for building

AI and CI has not been able to create decent human-AI and CI has not been able to create decent human-computer interfaces, solve problems in computer vision, computer interfaces, solve problems in computer vision, natural language understanding, cognitive search and data natural language understanding, cognitive search and data mining, or even reasoning in theorem proving. mining, or even reasoning in theorem proving.

Practical: humanized, cognitive computer applications Practical: humanized, cognitive computer applications require a brain-like architecture (either software or require a brain-like architecture (either software or hardware) to deal with such problems efficiently; it is at the hardware) to deal with such problems efficiently; it is at the center of cognitive robotics. center of cognitive robotics.

Some failed attempts• Many have proposed the construction of brain-like computers, Many have proposed the construction of brain-like computers,

frequently using special hardware. frequently using special hardware.

• Connection Machines from Thinking Machines, Inc. (D. Hills, Connection Machines from Thinking Machines, Inc. (D. Hills, 1987) was commercially almost successful, but never become 1987) was commercially almost successful, but never become massively parallel and the company went bankrupt.massively parallel and the company went bankrupt.

• CAM Brain (ATR Kyoto) – failed attempt to evolve the large-CAM Brain (ATR Kyoto) – failed attempt to evolve the large-

scale cellular neural network; based on a bad idea that one can scale cellular neural network; based on a bad idea that one can evolve functions without knowing them. It is impossible to evolve functions without knowing them. It is impossible to repeat evolutionary process (lack of data about initial organisms repeat evolutionary process (lack of data about initial organisms and environment, almost infinite number of evolutionary and environment, almost infinite number of evolutionary pathways). Evolutionary algorithms require supervision (fitness pathways). Evolutionary algorithms require supervision (fitness function) but it is not clear how to create fitness functions for function) but it is not clear how to create fitness functions for particular brain structures without knowing their functions first; particular brain structures without knowing their functions first; but if we know the function we can program it without evolving.but if we know the function we can program it without evolving.

Special hardware?• Many have proposed the construction of brain-like computers, Many have proposed the construction of brain-like computers,

frequently using special hardware, but there are no large-scale frequently using special hardware, but there are no large-scale constructions so far.constructions so far.

• Needed: elements based on spiking biological neurons and the Needed: elements based on spiking biological neurons and the

layered 2-D anatomy of mammalian cerebral cortex.layered 2-D anatomy of mammalian cerebral cortex.

• ALAVLSI, Attend-to-learn and learn-to-attend with analog VLSI, EU ALAVLSI, Attend-to-learn and learn-to-attend with analog VLSI, EU IST Consortium 2002-2005, Plymouth, ETH, Uni Berne, Siemens.IST Consortium 2002-2005, Plymouth, ETH, Uni Berne, Siemens.A general architecture for perceptual attention and learning based on A general architecture for perceptual attention and learning based on neuromorphic VLSI technology. neuromorphic VLSI technology.

Coherent motion + speech categorization, project ends in 2005.Coherent motion + speech categorization, project ends in 2005.

• P-RAM neurons, KCL? P-RAM neurons, KCL?

Natural perception

Spectrogram of speech: hearing a sentence.Spectrogram of speech: hearing a sentence.

Spiking vs. mean fieldBrain: 10Brain: 101111 Neurons Neurons

Neuron PoolsNeuron Pools

0

( , )( ) lim spikes

t

n t t tA t

M t

Pool Activity:

Pool

M Neurons

neuron

123

M

spikes t

t

Linked Pools Linked Pools

A

C

B

( ) ...B B extI I aF A It

Mean-FieldModel:

F

neuron 1neuron 1neuron 2neuron 2

Networks of Spiking NeuronsNetworks of Spiking Neurons

)())(()( tIVtVgtVdt

dsynLimim

2 ( )V t

t

Integrate andFire Model:

,, ,

,, ,

, ,, 2

, ,

( ) ( ( ) ) ( )

( ) ( ( ) ) ( )

( ( ) )( ) ( )

(1 [ ]exp( 0.062 ( ) /3.57))

( )

AMPA extAMPA ext AMPA ext i E ij j

j

AMPA recAMPA rec AMPA rec i E ij j

j

NMDA rec i E NMDA recNMDA rec ij j

ji

GABA rec GABA

I t g V t V w s t

I t g V t V w s t

g V t VI t w s t

Mg V t

I t g

,( ( ) ) ( )GABA rec

rec i E ij jj

V t V w s t

,

,

( )( ) ( )

( )( ) ( )(1 ( ))

( )( ) ( )

( )( ) ( )

AMPAjAMPA k

j jkAMPA

NMDAjNMDA NMDA

j j jNMDA decay

NMDAjNMDA k

j jkNMDA rise

GABAjGABA k

j jkGABA

s tds t t t

dt

s tds t x t s t

dt

x tdx t t t

dt

s tds t t t

dt

Synaptic Dynamics

EPSP, IPSP)(tI syn

Spike

Spike

SomaSynapses

synCmC mR

synR

Darwin/Nomad robotsG. Edelman G. Edelman (Neurosciences Institute) & collaborators, created a series (Neurosciences Institute) & collaborators, created a series of Darwin automata, brain-based devices, “physical devices whose of Darwin automata, brain-based devices, “physical devices whose behavior is controlled by a simulated nervous system”.behavior is controlled by a simulated nervous system”.

(i)(i) The device must engage in a behavioral task. The device must engage in a behavioral task.

(ii)(ii) The device’s behavior must be controlled by a simulated The device’s behavior must be controlled by a simulated nervous system having a design that reflects the brain’s nervous system having a design that reflects the brain’s architecture and dynamics. architecture and dynamics.

(iii)(iii) The device’s behavior is modified by a reward or value system that The device’s behavior is modified by a reward or value system that signals the salience of environmental cues to its nervous system.signals the salience of environmental cues to its nervous system.

(iv)(iv) The device must be situated in the real world.The device must be situated in the real world.

Darwin VII consists of: a mobile base equipped with a CCD camera Darwin VII consists of: a mobile base equipped with a CCD camera and IR sensor for vision, microphones for hearing, conductivity sensors and IR sensor for vision, microphones for hearing, conductivity sensors for taste, and effectors for movement of its base, of its head, and of a for taste, and effectors for movement of its base, of its head, and of a gripping manipulator having one degree-of-freedom; 53K mean firing gripping manipulator having one degree-of-freedom; 53K mean firing +phase neurons, 1.7 M synapses, 28 brain areas.+phase neurons, 1.7 M synapses, 28 brain areas.

Blue BrainThe Blue Brain Project was launched by the Brain Mind Institute, EPFL, The Blue Brain Project was launched by the Brain Mind Institute, EPFL, Switzerland and IBM, USA in MaySwitzerland and IBM, USA in May’’0505, now , now over 120'000over 120'000 WWW pages WWW pages..

The EPFL Blue Gene is the 8th fastest supercomputer in the worldThe EPFL Blue Gene is the 8th fastest supercomputer in the world..

CCan simulate about 100an simulate about 100M M minimal compartment neurons or 10-50'000 minimal compartment neurons or 10-50'000 multi-compartmental neuronsmulti-compartmental neurons, , with 10with 1033-10-1044 xx more synapses. more synapses. NNext ext generation generation BG will BG will simulate simulate >10>1099 neurons with significant complexity.neurons with significant complexity.

FFirst objective is to create a cellular level, software replica of the irst objective is to create a cellular level, software replica of the Neocortical Column for real-time simulations. Neocortical Column for real-time simulations.

The Blue Brain Project will soon invite researchers to build their own The Blue Brain Project will soon invite researchers to build their own models of different brain regions in different species and at different models of different brain regions in different species and at different levels of detail using Blue Brain Software for simulation on Blue Gene. levels of detail using Blue Brain Software for simulation on Blue Gene. These models will be deposited in an Internet Database from which These models will be deposited in an Internet Database from which Blue Brain software can extract and connect models together to build Blue Brain software can extract and connect models together to build brain regions and begin the first whole brain simulations. brain regions and begin the first whole brain simulations.

Blue Brain 2Models at different level of complexity: Models at different level of complexity: http://http://bluebrainproject.epfl.chbluebrainproject.epfl.ch//

1. The Blue Synapse: A molecular level model of a single synapse1. The Blue Synapse: A molecular level model of a single synapse..

2. The Blue Neuron: A molecular level model of a single neuron2. The Blue Neuron: A molecular level model of a single neuron..

3. The Blue Column: A cellular level model of the Neocortical column 3. The Blue Column: A cellular level model of the Neocortical column with 10with 10K K neuronsneurons, , later later 50K, 100M connections. 50K, 100M connections.

4. The Blue Neocortex: A simplified Blue Column will be duplicated to 4. The Blue Neocortex: A simplified Blue Column will be duplicated to produce Neocortical regions and eventually and entire Neocortex. produce Neocortical regions and eventually and entire Neocortex.

5. The Blue Brain Project will also build models of other Cortical and 5. The Blue Brain Project will also build models of other Cortical and Subcortical models of the brainSubcortical models of the brain, , and sensory + motor organsand sensory + motor organs..

Blue ColumnA detailed and faithful computer reproduction of the Neocortical Column. A detailed and faithful computer reproduction of the Neocortical Column.

It will first be based on the data obtained from rat somatosensory cortex It will first be based on the data obtained from rat somatosensory cortex at 2 weeks of age. Once built and calibrated with iterative simulations at 2 weeks of age. Once built and calibrated with iterative simulations and experiments, comparative data will be used to build columns in and experiments, comparative data will be used to build columns in different brain regions, ages and species, including humans. different brain regions, ages and species, including humans.

BC will be composed of 10BC will be composed of 1044 morphologically complex neurons with morphologically complex neurons with active ionic channels, interconnected in a 3-dimensional (3D) space active ionic channels, interconnected in a 3-dimensional (3D) space with 10with 1077 -10 -1088 dynamic synapses, receiving 10 dynamic synapses, receiving 1033 -10 -1044 external input external input synapses, generating 10synapses, generating 1033 -10 -1044 external output synapses. external output synapses.

Neurons use dynamic and stochastic synaptic transmission rules for Neurons use dynamic and stochastic synaptic transmission rules for learning, with meta-plasticity, supervised & reward learning algorithms learning, with meta-plasticity, supervised & reward learning algorithms for all synapses.for all synapses.

Blue Column 3Project will include creation of: Project will include creation of:

• Databases: NOBASE holds 3D reconstructed model neurons, Databases: NOBASE holds 3D reconstructed model neurons, synapses, synaptic pathways, microcircuit statistics, computer model synapses, synaptic pathways, microcircuit statistics, computer model neurons, virtual neurons. neurons, virtual neurons.

• Visualization: BlueBuilder, BlueVision and BlueAnalsysis. 2D, 3D Visualization: BlueBuilder, BlueVision and BlueAnalsysis. 2D, 3D and immersive visualization systems are being developed.and immersive visualization systems are being developed.

• Simulation: a simulation environment for large scale simulations of Simulation: a simulation environment for large scale simulations of morphologically complex neurons on 8000 processors of IBM's Blue morphologically complex neurons on 8000 processors of IBM's Blue Gene supercomputer. Gene supercomputer.

• Simulations & experiments: iterations between large scale Simulations & experiments: iterations between large scale simulations of neocortical microcircuits and experiments in order to simulations of neocortical microcircuits and experiments in order to verify the computational model and explore predictions.verify the computational model and explore predictions.

• Verification: in vivo = in silico? Verification: in vivo = in silico?

CCortex

Artificial Development (www.ad.com) is building CCortex™, Artificial Development (www.ad.com) is building CCortex™, a complete 20G neuron 20T connection simulation of the a complete 20G neuron 20T connection simulation of the Human Cortex and peripheral systems, on a cluster of 500 Human Cortex and peripheral systems, on a cluster of 500 computers - the largest neural network created to date.computers - the largest neural network created to date.

Artificial Development plans to deliver a wide range of commercial Artificial Development plans to deliver a wide range of commercial products based on artificial versions of the human brain that will products based on artificial versions of the human brain that will enhance business relationships globally. enhance business relationships globally.

Rather unlikely? Simulation of PentiumRather unlikely? Simulation of Pentium

Not much has changed in the last year on their web pageNot much has changed in the last year on their web page, , except that except that AD opened a lab in Kochi, Kerala, India, to “uncover relevant information AD opened a lab in Kochi, Kerala, India, to “uncover relevant information on the functioning on the human brain, and help model and interpret the on the functioning on the human brain, and help model and interpret the data.” The company is run by Marcos Guillen, who made money as ISP data.” The company is run by Marcos Guillen, who made money as ISP in Spain but has no experience in neuroscience or simulations.in Spain but has no experience in neuroscience or simulations.

The Ersatz Brain Project The Ersatz Brain Project

Vision: in 2050 the personal computer you buy in Wal-Mart will have two CPU’s Vision: in 2050 the personal computer you buy in Wal-Mart will have two CPU’s with very different architecture:with very different architecture:

First, a traditional von Neumann machine that runs spreadsheets, does First, a traditional von Neumann machine that runs spreadsheets, does word processing, keeps your calendar straight, etc. etc. word processing, keeps your calendar straight, etc. etc.   

Second, a brain-like chip Second, a brain-like chip

· To handle the interface with the von Neumann machine, · To handle the interface with the von Neumann machine,

·   Give you the data that you need from the Web or your files.·   Give you the data that you need from the Web or your files.

·   Be your silicon friend, guide, and confidant. ·   Be your silicon friend, guide, and confidant.

Project based on modeling of cortical columns of various sizes Project based on modeling of cortical columns of various sizes (minicolumns ~10(minicolumns ~1022, plain ~10, plain ~1044, and hypercolumns ~10, and hypercolumns ~1055), sparsely ), sparsely connected (0.001% in the brain). connected (0.001% in the brain).

NofN, Network of Networks approximation using 2D BSB (Brain in a Box) NofN, Network of Networks approximation using 2D BSB (Brain in a Box) network, similar in design to Connection Machines, but more processors.network, similar in design to Connection Machines, but more processors.

Conscious machines: HaikonenConscious machines: HaikonenConscious machines: HaikonenConscious machines: Haikonen

Haikonen has done some simulations based on a rather straightforward Haikonen has done some simulations based on a rather straightforward design, with neural models feeding the sensory information (with WTA design, with neural models feeding the sensory information (with WTA associative memory) into the associative “working memory” circuits. associative memory) into the associative “working memory” circuits.

by Tetsuya HoyaBSI-RIKEN, Japan

Lab. Advanced Brain Signal Processing

Artificial Mind System (AMS)Artificial Mind System (AMS)Kernel Memory ApproachKernel Memory Approach

Series: Studies in Computational Intelligence (SCI), Vol. 1 (270p)

Springer-Verlag: HeidelbergAug. 2005

available from: http://www.springeronline.com/

• To provide an engineering account to model various functionalities related to mind, motivated from the modularity principle of mind (Fodor, 1983; Hobson, 1999).

• To embody each module and their mutual data processing within the AMS, by means of a new connectionist model, kernel memory.

• Thereby, to develop a new form of artificial intelligent system with ideas from a broader spectrum of brain scientific studies – artificial intelligence, cognitive science/psychology, connectionism, consciousness studies, general neuroscience, linguistics, pattern recognition/data clustering, robotics, and signal processing.

Objectives:

Artificial Mind System (AMS)Artificial Mind System (AMS)Kernel Memory ApproachKernel Memory Approach

Machine consciousness: OwenMachine consciousness: Owen

Holland Owen, ExeterHolland Owen, Exeterhttp://www.machineconsciousness.org/http://www.machineconsciousness.org/

Owen Holland at the University of Essex and Tom Troscianko and Ian Owen Holland at the University of Essex and Tom Troscianko and Ian Gilchrist at the University of Bristol, have received £493,000 (714,000 Gilchrist at the University of Bristol, have received £493,000 (714,000 Euros, or $833,000) from the Eng. & Phys. Sci. Res. Council for a Euros, or $833,000) from the Eng. & Phys. Sci. Res. Council for a project 'Machine consciousness through internal modeling‘, 2004-2007. project 'Machine consciousness through internal modeling‘, 2004-2007.

To survive robots will plan actions, build a model of the world and a To survive robots will plan actions, build a model of the world and a model of itself - its body, sensors, manipulators, preferences, history … model of itself - its body, sensors, manipulators, preferences, history … Biological vision systems is the basis for internal processes and models Biological vision systems is the basis for internal processes and models and will be accessible to the investigating team as visual displays. The and will be accessible to the investigating team as visual displays. The main focus of interest will be the self-model; its characteristics and main focus of interest will be the self-model; its characteristics and internal changes are expected to resemble those of the conscious self internal changes are expected to resemble those of the conscious self in humans, perhaps closely enough to enable some of the robots to be in humans, perhaps closely enough to enable some of the robots to be regarded as possessing a form of machine consciousness.regarded as possessing a form of machine consciousness.Increasingly complex biologically inspired autonomous mobile robots Increasingly complex biologically inspired autonomous mobile robots forced to survive in a series of progressively more difficult forced to survive in a series of progressively more difficult environments, and will then study the external and internal behavior of environments, and will then study the external and internal behavior of the robots, looking for signs and characteristics of consciousness. the robots, looking for signs and characteristics of consciousness.

Bayesian Confidence Propagating NN.Bayesian Confidence Propagating NN.

Johansson/Lansner ideas:Johansson/Lansner ideas:

o Assumption: functional principles of cortex reside on a much higher Assumption: functional principles of cortex reside on a much higher level of abstraction than that of the single neuron i.e. closer to level of abstraction than that of the single neuron i.e. closer to abstractions like ANN and connectionist models.abstractions like ANN and connectionist models.

o Target: artificial brain, compact, low-power, multi-network NN. Target: artificial brain, compact, low-power, multi-network NN. o Mapping of cortical structure onto the BCPNN, an attractor network.Mapping of cortical structure onto the BCPNN, an attractor network.o Implementation of BCPNN based on hyper columnar modules.Implementation of BCPNN based on hyper columnar modules.

o Hypercolumn needs 5Hypercolumn needs 5..101099 ops, with about 2 ops, with about 2..101066 hypercolumns in hypercolumns in human cortex, giving about 10human cortex, giving about 101616 ops. ops.

o No detailed structure proposed.No detailed structure proposed.

Intelligent Distributed Agents.Intelligent Distributed Agents.

Stan Franklin (Memphis): IDA is an intelligent, autonomous software Stan Franklin (Memphis): IDA is an intelligent, autonomous software agent that does personnel work for the US Navy. agent that does personnel work for the US Navy.

IDA insideIDA inside

Based on Baars “Global Workspace” theory. Based on Baars “Global Workspace” theory.

IDA in actionIDA in action

Hal Baby BrainHal Baby Brain..

Evolve language: Evolve language: www.a-i.comwww.a-i.com

So far: simple 2-3 words but meaningful.So far: simple 2-3 words but meaningful.

Will it ever make it to higher level? Doubtful. Will it ever make it to higher level? Doubtful.