WHY ROBOTICS, AI, AL & QUANTUM COMPUTING
4 likes694 views
The document discusses the complexities and challenges of understanding the universe, technology, and intelligence, emphasizing the importance of embracing advanced technologies like AI, quantum computing, and robotics for future progress. It highlights the limitations of current models in various fields such as economics, biology, and ecology, while advocating for a deeper understanding and application of these technologies to solve significant problems. Furthermore, the text concludes that intelligence is an emergent property of complex systems, and the cooperation between AI and humans is crucial for advancement in capabilities and innovation.
![Ic ~ k log2[1 + K.A.S ( 1 + P.M )]
S = Sensor, A = Actuator, P = Processor, M = Memory
I N T E L L I G E N C E D E F I N I T I O N](https://image.slidesharecdn.com/iodessexuniversity19sept2019-190923110811/85/WHY-ROBOTICS-AI-AL-QUANTUM-COMPUTING-49-320.jpg)
WHY ROBOTICS, AI, AL & QUANTUM COMPUTING
- 1. Why Robots, AI, AL AND Quantum Computing ? Prof Peter Cochrane OBE petercochrane.com Sentient Systems
- 2. Innovation Centre 19 Sept 2019
- 3. REALITY STRIKES The universe is not simple or linear The universe is complex, non-linear and chaotic Workable simplicity is long gone - complexity now rules ! Robert Boyle PV = MRT George Ohm V = IR Robert Hook x = Fk Isaac Newton F = Mg
- 4. REALITY STRIKES - Thank You Human Limitations! Simple thinking has seen us achieve so much! Further progress however demands better models and deeper understanding “Just because we don’t understand something does not mean to say we cannot exploit it”
- 5. IGNORANCE - What we do not understand Economics - the current model is destroying the planet Climate Change - if we do nothing it might just destroy us Biology - we do not understand cells or stem cells sufficiently well Ecologies - we are relatively clueless and continue to inflict great damage Chemistry - our models are crude and preclude the synthesis of the new Physics - quantum mechanics + string theory create more paradoxes than they solve Networks - we design and build best we can but are unable to predict outcomes Software - we have no means of thoroughly testing what we build People - behaviours remain something of a mystery +++++ - “Our mathematical models/abilities are virtually useless at the leading edge providing nothing more than crude representations at best ”
- 6. HOPE - What might just save us ? 260 × 280 Deep analysis and modelling Deep cognition and knowledge
- 7. OUR CHOICE - Binary! We can pretend we don’t need new technologies and thereby suffer a slow death at the hand of fate…. …or we can embrace and exploit to our advantage whilst discovering the problems and dangers - and then resolve, refine and repair as we go… We actually have no choice if we wish to survive!
- 8. DON’t Panic - Think Instead ! Naysayers in General never built an AI system or studied the topic
- 9. Reality CHECK - The Real Upside(s) More accurate medical diagnostics Acceleration of genomic medicine Highly efficient medical imaging Combinatorial drug analysis New drug discovery New materials Network analysis Dynamic modelling Automated security Solving problems beyond human intellect Technology is inert and does not build dystopian societies - only people do that! Robots and AI rapidly learn from humans
- 10. BIO-TECH nano-TECH AIRoboticsMulti-Disciplinary hot spot for the 21C New materials New industries New processes New capabilities Lower energy Lower waste Less friction The IoT SUSTAINABILTY
- 11. AWARENESS Begets sentience A function of context and cognition An indirect function of memory & computing power Predominantly determined by sensors, actuators and data I/O Suppositions All living things exhibit intelligence <=> Not all intelligent things exhibit life Humans do not have the capacity to recognise all intelligences Not all intelligent/living systems enjoy/exhibit awareness
- 12. RECAP - Reality Check Reasonable Deterministic and Well Behaved
- 13. RECAP - Reality Check Unreasonably Non-Linear and Chaotic Huh !!
- 14. BIO-LIFE
- 15. ARTIFICIAL LIFE - Breeding Software Huh !!
- 16. Robots - Status Quo Power industry, farming, production and delivery, and subliminally feature in most of our lives They do a better job than humans by working to a far greater precision, not getting bored and not losing concentration
- 17. Robots - Status Quo The next phase: buddy-buddy working ! Human - Robot : Robot - Human helpers/teams !
- 18. DOING WHAT WE DO - Embedded AI Learning
- 19. DOING WHAT WE DO - Embedded AI Learning
- 20. DOING WHAT WE Cannot DO - Inside Us ! Productionizing surgery and medical procedures
- 21. Are you an artificial intelligence ? Does it Matter ? Future reality Already here - a widely accepted norm Talking to your car, TV, computer mobile - Amazon Alexa et al are examples of very limited and weak apps/specific AI General AI is much harder to crack - but is being realised a byte at a time…
- 22. 7 0 y e a r s o n ! “We can assert with great confidence that we do not yet fully understand the human (or any other) brain” This is not going to get any bigger, better, or in anyway more powerful and it is a fundamental limiter to our further progress
- 23. C RU D e ! Neuron Count Early machines had ~4k memory & << 1k nodes Such comparisons are engineering estimates that are virtually useless and certainly not valid or accurate….
- 24. BIO-ENGINEERING Ratio of neurons + axon + dendrite network to vascular plumbing to get energy in and heat out STILL BEING WORKED ON: Physiology of men and women drastically different in respect of nerve endings across the surface are of the skin - and the eyes - for sure they are ‘wired’ differently! For average human adult "The average number of neocortical neurons was 19 Bn in female brains and 23 Bn in male brains." T 0 b e c o n t in u e d
- 25. 1949 DONALD HEBB Learning We remember, forget, learn, make decisions on the basis of the concatenation of accumulating switching and/or decaying synapse states as they store ‘charge’ almost capacitor like and thereby ‘influence’ their neighbours Forgetting appears to be essential for learning
- 26. Thermodynamics tells us that one of these cannot decode or understand the other We can only understand our own brain with the aid of something more powerful AI + Quantum Computing + Human ? Physics gets in on the act
- 27. ABILITY BENCHMARK 1031 v 959 cells - HUH ? When AI (/Quantum Computing ?) helps us crack this problem we will have the first comparative measure of what it might take to understand ourselves ! We ~1011 neurons cannot understand ~102 Worms Social behaviours with 302 neurons
- 28. 1957: 13 people deliver a computer 2017: 13 computers in one hand £29,000 then ~£613,000 today HD 320k >3.5kW Q u a d C o r e Memory 1.G 12W £20 S E G u e B A C K
- 29. i n < 3 0 Y e a r s ~ 1,000,000,000 x chip capacity Cray 2 1985 $32M and 5kW iPhone 5 2012 $700 and 5W 3 x more powerful
- 30. i n < 3 0 Y e a r s ~ 1,000,000,000 x chip capacity ~Dog Brain ~Mouse Brain ~Human Brain There is far more to intelligence than a very crude analogy to transistor - neuron equivalence and count, but this serves to indicate one reason why AI has been along time coming ! Processing Memory Sensors Adaptability SoftwareComplexity Autonomy AI Processing Memory Sensors Manipulators Communication Networking Intel > 100M transistors/mm2 IBM > 30Bn transistors/chip Feature size now ~ 10nm
- 31. G A M E C H A N G E R 2 0 1 4 Considered to be impossible by philosophers
- 32. G A M E C H A N G E R 2 0 1 8 Considered to be impossible by philosophers
- 33. Processing Speed Data Storage Internet Processing Indeterminate I n t e r n e t Storage ~1022 Bytes Human Storage ~1016 Bytes Processing ~1 TF Cat Storage ~1014 Bytes Processing ~0.1 TF T E C H V B I O L G Y A very rough comparison - the actuality is far more complex Super Computer Storage ~1017 Bytes Processing ~200 PetaFlops
- 34. Processing Speed Data Storage Internet Processing Indeterminate I n t e r n e t Storage ~1022 Bytes Human Storage ~1016 Bytes Processing ~1 TF Cat Storage ~1014 Bytes Processing ~0.1 TF T E C H V B I O L G Y A very rough comparison - the actuality is far more complex Super Computer Storage ~1017 Bytes Processing ~200 PetaFlops STILL Relatively D U M B c o m p a r e d to biology
- 35. RECOGNITION
- 36. Data Algorithm AnswerDigital Computer Data Exemplars AnswerAI (Neural Net Pattern Matching) Tuning Human Assisted Refinement Data Answer AlgorithmMachine Learning H m m ! Just to be clear…
- 37. WATC H T h is N ET
- 38. WATC H T h is N ET
- 39. A I Sense, Garner Data, Analyse, Reason M A C H I N E L e a r n i n g Adaptive Networks & Algorithms More exposure to data sees continual improvement in accuracy with time P E RS P ECT IV E D e e p L e a r n i n g Multilayered Neural Networks Vast amounts of data result in an ever more accurate consensus Can get very complex but I think we can safely say that we fully understand this Designed, Built, Optimised by Man Designed By Man & Self Optimising Whilst we understand most of this - there is a growing % of Huh ! Designed By Man & Machine Emergent behaviours dominate - we may or may not comprehend Boundaries not well defined
- 40. A I Sense, Garner Data, Analyse, Reason M A C H I N E L e a r n i n g Adaptive Networks & Algorithms More exposure to data sees continual improvement in accuracy with time P E RS P ECT IV E D e e p L e a r n i n g Multilayered Neural Networks Vast amounts of data result in an ever more accurate consensus Can get very complex but I think we can safely say that we fully understand this Designed, Built, Optimised by Man Designed By Man & Self Optimising Whilst we understand most of this - there is a growing % of Huh ! Designed By Man & Machine Emergent behaviours dominate - we may or may not comprehend Boundaries not well defined
- 41. G a m e C h a n g e r 2 o 1 8 Goodbye Touring Test and a lot of nonsense P re - p ro g ra m m e d c o n t e x t a n d o n l y limited cogn ition required !
- 42. G a m e C h a n g e r 2 o 1 8 Goodbye Touring Test and a lot of nonsense T h e m a c h i n e i s d e f i n i t e l y s m a r t e r t h a n t h e h u m a n i n t h i s c a s e !
- 43. A I C O N T E X T A vital step toward cognition Feel Sight Taste Smell Sound Experiences Cameras Mobiles Sensors Media Nets IoT “Let us pray” “Lettuce spray” “Let us spray” Are we in church/mosque, kitchen or garden ? Ultimately, AI is defined by sensory systems
- 44. AWAREness More & more clues Who is it What are they Where are they What is the topic A priori information Content & context Facial expressions Body language All combat noise & other generators of errors
- 45. AWAREness The military version! Hardly likely to be sufficient (or work) for modern non-linear warfare - probably as rigid and as smart as it appears
- 46. 5 TYPE 1 Reactive Task Specific Very Limited Largely Pattern Matching Human Programmers: Chess, cards, dominoes data, speech, pictures, characters, behaviours, movements+ Narrow Cognition Largely Programmed by AI alone: Subsume the networked knowledge of previous and current generations TYPE 3 Reasoning Multi-Task Ability Broadly Applicable TYPE 2 Learning Task Specific Broadly Applicable Memory and Analysis Initial Human AI Program That Then Adapts: Recognises highly complex/ large scale non-linear relationships A I L A D D E R Progress & Categories Full Awareness May be categorised as a ‘being’: With a wide range of sensory units networked to other machines TYPE 4 Self-Aware
- 47. 5 TYPE 1 Reactive Task Specific Very Limited Largely Pattern Matching Human Programmers: Chess, cards, dominoes data, speech, pictures, characters, behaviours, movements+ Narrow Cognition Largely Programmed by AI alone: Subsume the networked knowledge of previous and current generations TYPE 3 Reasoning Multi-Task Ability Broadly Applicable TYPE 2 Learning Task Specific Broadly Applicable Memory and Analysis Initial Human AI Program That Then Adapts: Recognises highly complex/ large scale non-linear relationships A I L A D D E R Progress & Categories Full Awareness May be categorised as a ‘being’: With a wide range of sensory units networked to other machines TYPE 4 Self-Aware ~ 70yrs to become a solid & deployable technology 2027 Robotic embodiment, extensive sensors & actuators plus entity networking rapidly raises the game 2017 Google Alpha GO p r o p e l s A I i n t o a n autonomous future of learning & doing 2024 AI-Human cooperation see exponential innovation & progress of AI capabilities
- 48. P R O G ESS S t i l l a t t h e R & D s t a g e A I R o b ots Q C D e p l o y m e n t f a r f a s t e r t h a n R o b o t i c s A L
- 49. Ic ~ k log2[1 + K.A.S ( 1 + P.M )] S = Sensor, A = Actuator, P = Processor, M = Memory I N T E L L I G E N C E D E F I N I T I O N
- 50. 1960 70 80 90 2000 10 ComputingPowerMIP/s Intelligence Level PC Estimates of Artificial Intelligence Internet R i s e o f I n t e l l i g e n c e
- 51. Ex e m p l a r N e w C r e a t i v i t y Criticism It is just copying and aping what human composers have done ! Retort 1 You mean exactly like human composers do ? Retort 2 AI has only just got into this game that humans have be at for well over 3M years
- 52. Ex e m p l a r N e w C r e a t i v i t y
- 53. Thinking the right way People fear, & complain, overlook the big gains V i v e L a d i f f e r e n c e
- 54. WHAT HAVE WE DISCOVERED Intelligence is an emergent property of complex systems Not all complex systems exhibit intelligence Intelligence always involves communication All intelligences are entropic AI always seeks the truth Axioms An intelligent system must have some form of input and an output Memory and processing power are not always an essential Communication and connection are essentials Things that think want to link Things that link want to think
- 55. ONLY ONE All that is needed Mobile AI/Robotic weapon platforms demand far less resources than any atom bomb and/or missile program Numerous rogue states will quickly embrace this new opportunity! W e a p o n s C o n t r o l S E E M S I m p o s s i b l e
- 56. TOASTER Deep humour or nightmare? Talkie Toaster, in Red Dwarf novels & TV series. Manufactured by Crapola Inc., an annoying, monomaniacal, AI toaster purchased by Dave Lister whilst on leave at a second-hand junk shop on Miranda, along with a robot goldfish and a smuggled pet cat. More intelligent than the Red Dwarf computer Holly
- 57. Thank You petercochrane.com What are these crazy humans planning to do next ? Sentient Systems