What do machines think
0 likes376 views
The document explores the nature of intelligence, questioning how machines think and the comparative efficacy of complex systems. It concludes that intelligence is logarithmic rather than linear, highlighting our dependency on technology and the challenges of defining intelligence. The author argues for a deeper understanding of intelligent systems and their role in addressing complex societal and environmental issues.
![The answer !
Ic ~ k log2[1 + KAS ( 1 + PM )]
S = Sensor, A = Actuator, P = Processor, M = Memory
Relative Intelligence = Ic
Flops (or machine cycles)
Observations:
With zero processor power intelligence is still possible
With zero memory power intelligence is still possible
With zero sensor and/or actuator power intelligence is impossible](https://image.slidesharecdn.com/whatdomachinesthink-101111093652-phpapp01/85/What-do-machines-think-4-320.jpg)
![And for a very simplistic model where
normalise complex operators, and
combine processing and memory types
we deduce that...
Ic = I Ei - Eo I = I Ei - E[a.s ( 1 + m + p + mp ) ] I](https://image.slidesharecdn.com/whatdomachinesthink-101111093652-phpapp01/85/What-do-machines-think-117-320.jpg)
![And for a very simplistic model where
normalise complex operators, and
combine processing and memory types
we deduce that...
Ic = I Ei - Eo I = I Ei - E[a.s ( 1 + m + p + mp ) ] I
Where: a, p ,m, s are all complex operators](https://image.slidesharecdn.com/whatdomachinesthink-101111093652-phpapp01/85/What-do-machines-think-118-320.jpg)
![After some mathematical
juggling and normalisation,
liberties, crimes, & of course
engineering licence...
Processor Memory
∝ log [1 + S. A (1 + P + M + P.M ) ]
I 2
Sensor Actuator](https://image.slidesharecdn.com/whatdomachinesthink-101111093652-phpapp01/85/What-do-machines-think-119-320.jpg)
![And finally, something we can work with:
Ic = k log2 [1 + K.S.A (1 + P + M + P.M ) ]
System Constants
For P.M >> (P + M)
Ic ~ k log2 [1 + K.S.A (1 + P.M ) ]
Relative Intelligence k log2 [1 + K.S.A (1 + P.M ) ]
=
of different systems Flops (or machine cycles)](https://image.slidesharecdn.com/whatdomachinesthink-101111093652-phpapp01/85/What-do-machines-think-120-320.jpg)
What do machines think
- 1. What Do Machines Think? Peter Cochrane www.ca-global.org COCHRANE a s s o c i a t e s
- 2. The problem ? * How do you compare complex man made systems ? * How do you compare solution efficacy ? * How do you minimise risk, maximise success ? * Mission critical - lives at risk - we have to move fast!
- 3. The key questions ? * What is intelligence ? * Can it be quantified ?
- 4. The answer ! Ic ~ k log2[1 + KAS ( 1 + PM )] S = Sensor, A = Actuator, P = Processor, M = Memory Relative Intelligence = Ic Flops (or machine cycles) Observations: With zero processor power intelligence is still possible With zero memory power intelligence is still possible With zero sensor and/or actuator power intelligence is impossible
- 5. Key conclusion: Intelligence grows logarithmically and not linearly ! The singularity event isn’t going to happen just yet ! The assumption that is is is based on I ~ PM growth looks to be wrong Kurzweil et al
- 6. Two wisdoms from my past ! “Whilst it is acceptable for the mathematicians, physicists and theorists to declare that there is no solution, we in engineering enjoy no such luxury. We always have to find a solution” ! “Before you even start a problem it is worth thinking what form the answer might be, and what would be reasonable”
- 7. Where did my thinking start ?
- 8. Where did my thinking start ? …in a Spanish gutter/drain!
- 9. Where did my thinking start ? …in a Spanish gutter/drain!
- 10. Where did my thinking start ? …in a Spanish gutter/drain!
- 11. Where did my thinking start ? …in a Spanish gutter/drain! …plus amateur robotics!
- 12. Where did my thinking start ? …in a Spanish gutter/drain! …plus amateur robotics!
- 13. Where did my thinking start ? …in a Spanish gutter/drain! …plus amateur robotics!
- 14. And then I got into the logistics of people, goods and traffic...and the traveling salesman problem...
- 15. ...and so I progressed to ‘wet science’ with genetics & sexual reproduction migrating into machines...
- 16. SO, I’m going to ask you to suspend all preconceptions, and beliefs...
- 17. SO, I’m going to ask you to suspend all preconceptions, and beliefs... Because we have to take the wider view, to be receptive...as we are mostly flying blind... ...although there is a growing body of fundamental knowledge and evidence...
- 18. Science fiction & Hollywood are mostly apocalyptic on the topic:
- 19. Science fiction & Hollywood are mostly apocalyptic on the topic:
- 20. Science fiction & Hollywood are mostly apocalyptic on the topic:
- 21. Science fiction & Hollywood are mostly apocalyptic on the topic:
- 22. Science fiction & Hollywood are mostly apocalyptic on the topic: …but the reality is quite different..so far!
- 23. Azimov’s Laws or Robotics Zeroth Law (1985): A robot may not injure humanity, or, through inaction, allow humanity to come to harm First Law (1940): A robot may not injure a human being, or, through inaction, allow a human being to come to harm, unless this would violate the Zeroth Law of Robotics Second Law (1940): A robot must obey orders given it by human beings, except where such orders would conflict with the Zeroth or First Law Third Law (1940): A robot must protect its own existence as long as such protection does not conflict with the Zeroth, First, or Second Law
- 24. Azimov’s Laws or Robotics Zeroth Law (1985): A robot may not injure humanity, or, through inaction, allow humanity to come to harm First Law (1940): We have chosen not to invoke these laws A robot may not injure a human being, or, through inaction, allow a human being to come to harm, unless this would violate the Zeroth Law of Robotics Second Law (1940): We have also opted to violate them A robot must obey orders given it by human beings, except where such orders would conflict with the Zeroth or First Law Third Law (1940): A robot must protect its own existence as long as such protection does not conflict with the Zeroth, First, or Second Law
- 25. Some realities: GOOD NOT SO GOOD
- 26. Some realities: GOOD NOT SO GOOD
- 27. Some realities: GOOD NOT SO GOOD
- 28. Some realities: GOOD NOT SO GOOD
- 29. Some realities: GOOD NOT SO GOOD
- 30. Some realities: GOOD NOT SO GOOD
- 31. Some realities: GOOD NOT SO GOOD
- 32. Some realities: GOOD NOT SO GOOD
- 33. Some realities: GOOD NOT SO GOOD
- 34. Some realities: GOOD NOT SO GOOD
- 35. Some realities: GOOD NOT SO GOOD
- 36. Some realities: GOOD NOT SO GOOD …so far we have been in control....!
- 37. The big question = Is there a big upside ? Or even a silver lining ? YES - one BIG example now and a few more later
- 38. Futures Advance Society Technology Wealth Health Well Being Time
- 39. Futures Advance Society Technology Wealth Health Well Being Oil runs out We do nothing Planet retaliates Life dies Time
- 40. Futures Advance Society Technology Wealth Our green effort works Health & status quo just holds Well Being Oil runs out We do nothing Planet retaliates Life dies Time
- 41. Cold Fusion Futures Climatic Control Advance Earth Management Nano - Bio Tech Society AI + AL +++ Technology Wealth Our green effort works Health & status quo just holds Well Being Oil runs out We do nothing Planet retaliates Life dies Time
- 42. Cold Fusion Futures Climatic Control Advance Earth Management Nano - Bio Tech Society AI + AL +++ Technology Our climate and planet models are Wealth Our green effort works so crude that we have&no real clue Health status quo just holds about which trajectory we are on Well Being …..and if we can influence itout Oil runs anyway! We do nothing Planet retaliates Life dies Time
- 43. Everything in the universe is governed by non-linearity and chaotic actions….
- 44. Everything in the universe is governed by non-linearity and chaotic actions…. …but we are linear and limited thinkers devoid of tools & techniques to help us comprehend…
- 45. Everything in the universe is governed by non-linearity and chaotic actions…. …but we are linear and limited thinkers devoid of tools & techniques to help us comprehend… If we have a long-term future it lies with our technology and machines and their ability to extend our knowledge & understanding.
- 46. If we switched off our machines today….
- 47. If we switched off our machines today…. …this lonely isolated planet would support < 1 Bn people
- 48. If we switched off our machines today…. …this lonely isolated planet would support < 1 Bn people Our total dependence on technology for the survival of our species is long established & will not go away, in fact, it continues to grow!
- 49. No human leader, past and present, future has the wisdom & ability to govern a chaotic planet!
- 50. No human scientist, past and present, future has the wisdom & ability to understand a chaotic planet!
- 51. Hypothesis The combined intelligence of all humanity is also insufficient
- 52. Hypothesis Our only hope is The combined the combined intelligence of intelligence all humanity is of also insufficient our machines
- 54. Definition - Think? • Judge or regard • Imagine or visualise • Mental preoccupation • Expect, believe, suppose • Have in mind as a purpose • Make inferences, decisions • Reflect on, or reason about • Decide by reasoning/reflecting • Exercise one's power of reason • Recall knowledge from memory • Arrive at a solution or judgments • Be capable of conscious thought • Dispose the mind in a certain way • Bring into a given mental condition
- 55. Definition - Think? • Judge or regard • Imagine or visualise • Mental preoccupation • Expect, believe, suppose • Have in mind as a purpose • Make inferences, decisions • Reflect on, or reason about • Decide by reasoning/reflecting • Exercise one's power of reason • Recall knowledge from memory • Arrive at a solution or judgments • Be capable of conscious thought • Dispose the mind in a certain way • Bring into a given mental condition
- 56. Definition - Think? • Judge or regard • Imagine or visualise ?? • Mental preoccupation • Expect, believe, suppose ?? • Have in mind as a purpose • Make inferences, decisions • Reflect on, or reason about • Decide by reasoning/reflecting • Exercise one's power of reason • Recall knowledge from memory • Arrive at a solution or judgments • Be capable of conscious thought ?? • Dispose the mind in a certain way ?? • Bring into a given mental condition
- 57. What do we know? Thinking is an electro-chemical process... ...influenced by experience and memory, emotions, external and internal conditions, plus sensory system, actions and inaction Other ‘thinking life forms’ are considerably simpler!
- 58. Words that lack any full or adequate description, definition, quantification, measure or meaning...
- 59. Words that lack any full or adequate description, definition, quantification, measure or meaning... • Life • Intelligence • Complexity • Scaleability • Value • Think(ing)
- 60. Without adequate description, definition, quantification, then meaningful discussion is very limited & we soon descend into the realms of belief and folk law + superstition... So far our only successful route to understanding has been the application of the scientific method
- 61. Not fully understanding something doesn’t imply we cannot exploit it!
- 62. Not fully understanding something doesn’t imply we cannot exploit it! Ballistics Newton Einstein Linear Non-Linear
- 63. The scientific method is the most powerful tool of understanding we have developed so far… ..but it never does, and will never do this!
- 64. Do machines think good or bad thoughts? We have empowered them to do both We are responsible for their initiation - their starting point Evolution will take care of history
- 65. Our view of technology…
- 66. Our view of technology… My Grandfather: Man will never fly
- 67. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon
- 68. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon My Generation: Computers will never:
- 69. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon My Generation: Computers will never: - win at chess
- 70. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon My Generation: Computers will never: - win at chess - be creative
- 71. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon My Generation: Computers will never: - win at chess - be creative - smarter than us
- 72. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon My Generation: Computers will never: - win at chess - be creative - smarter than us - become intelligent
- 73. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon My Generation: Computers will never: - win at chess - be creative - smarter than us - become intelligent - become self aware
- 74. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon My Generation: Computers will never: - win at chess - be creative - smarter than us - become intelligent - become self aware - live, etc
- 75. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon My Generation: Computers will never: - win at chess - be creative - smarter than us - become intelligent - become self aware - live, etc My View: These are all worthless & irrelevant statements
- 76. Our view of technology… My Grandfather: Man will never fly My Father: Man will never walk on the moon My Generation: Computers will never: - win at chess - be creative - smarter than us - become intelligent - become self aware - live, etc My View: These are all worthless & irrelevant statements mainly born of fear and ignorance
- 77. Garry Kasporov...1997 Machines (IBM Big Blue) don’t play chess like us....
- 78. Garry Kasporov...1997 Machines (IBM Big Blue) don’t play chess like us.... ...and this is precisely why machines can add so much value....and understanding!
- 79. We now live in a world of the infinitely complex… … how do you solve problems & make decisions?
- 80. We now live in a world of the infinite… … how do you find and how do you choose?
- 81. No one knows anything anymore... highly specialised becoming all knowledge the fount specialised through of all knowledge technology 1800 1900 TODAY 2000 +
- 82. We have come to the end(ish) of the road… … and we aint going to get any smarter!
- 83. Viagra & chips for the brain
- 84. The only solutions on the horizon: A symbiotic relationship with our machines? A new intelligence/s and/or life form/s?
- 85. Some of us see the route to AI & AL through the direct imitation of mother nature & her already solved problems…
- 86. Some of us see the route to AI & AL through the direct imitation of mother nature & her already solved problems…
- 87. And then there are… …the hybrid solutions .…combining man and machine
- 88. Why do we see see technology as a threat… …and remain complacent about this stuff?
- 89. After 1.8Bn years we still can’t… …adequately describe, define, quantify or understand intelligence
- 90. And despite the dire warnings about the limitations of his test and the way it can be applied it is still used today as if it were actually useful!
- 91. Today ! IQ = Solving problems Power = Connectivity + neuron count Speed = Task completion
- 92. Today ! IQ = Solving problems Power = Connectivity + neuron count Speed = Task completion
- 93. Sometimes we have to make giant conceptual, and questionable leaps to get to a solution... ...and better to get a good approximation, or even the wrong answer, than no answer at all !
- 94. A valuable engineering precept: If there is no theor y and proof, or no understanding - any solution will do - so long as we make a start...
- 95. Approach
- 96. I start by doing a Dr Johnson and define intelligence! “And intelligence is the way I define it and no one else” Samuel Johnson (1709-1784) Johnson was an English writer and critic, and My misquote! one of the most famous literary figures of the 18th century. His best-known work is his 'Dictionary of the English Language
- 97. And I don’t mean wordy descriptions of the form... “The faculty of rational behaviour” “An ability to create order out of chaos”
- 98. Also, recognise that we often have difficulty discerning what is intelligent and what isn’t! Random scatter Wind action Wind action ? Very unlikely ? for sure ? or a Intelligent action most likely leaf blower ?
- 99. Also, recognise that we often have difficulty discerning what is intelligent and what isn’t! Observation: Intelligent intervention can prevent or even reverse the natural process of entropy within a system. All natural systems seek to find the lowest energy state - ie they migrate from order to disorder - the universe started as a point function and then expanded Random scatter Wind action Wind action ? Very unlikely ? for sure ? is the celestial ratchet - it only ever gets Entropy or a Intelligent action most likely bigger - and is leaf blower ? the mechanism by which we die!
- 100. This could never occur by random action, it definitely involves intelligence!
- 101. We are actually dealing the transformation of information ... Intelligence can thus be taken (Because I say so!) as the efficiency and efficacy by which that process is completed. This automatically infers an entropic quality: Intelligence = Entropy of Input - Entropy of Output & Efficacy of Course
- 102. This would not be a good answer ! Answer to the Ultimate Question of Life, the Universe, and Everything from the supercomputer, Deep Thought, specially built for this purpose. It takes Deep Thought 7½ million years to compute and check the answer, which turns out to be 42. Unfortunately, The Ultimate Question itself is unknown. Douglas Adams: The Hitchhikers Guide to The Universe
- 103. Because intelligent systems can both increase or decrease the total Information/Entropy, we define...
- 104. Because intelligent systems can both increase or decrease the total Information/Entropy, we define... Intelligence = I = MOD { Ei - Eo } = I Ei - Eo I
- 105. Because intelligent systems can both increase or decrease the total Information/Entropy, we define... Intelligence = I = MOD { Ei - Eo } = I Ei - Eo I Where the Entropy E = The amount of information required to exactly define the state of a system
- 106. At a very basic level an intelligent system could take the form... Sensors Processor S P1 System Memory Environs M Actuator Processor A P2
- 107. But recognise that they can often be a whole lot more complex... Processor Sensors S1 P1 Memory Mlong Memory System Environs M Memory short Mlocal Actuator Processor A1 P1 A
- 108. Eyes of all kinds often do a lot of pre processing
- 109. And you don’t put all the brains in the head of a dog or a human!
- 110. Back to the plot!
- 111. This is a tough enough problem for now! Sensors Processor S P1 System Memory Environs M Actuator Processor A P2
- 112. Assuming the following nomenclature.... si p1 si Sensors . Processor P1 . i S System Memory Environs M Actuator Processor (1 + m) p1 si o A P2 a(1 + m + p1 p2 + p2 m p1 ) si (1 + m + p1 p2 + p2 m p1 ) si Where: s, p ,m, a are complex operators
- 113. And with our very simplistic model we deduce that...
- 114. And with our very simplistic model we deduce that... Ic = I Ei - Eo I = I Ei - E( a[(1 + m + p1 p2 + p2 m p1 ) si I
- 115. And with our very simplistic model we deduce that... Ic = I Ei - Eo I = I Ei - E( a[(1 + m + p1 p2 + p2 m p1 ) si I Where: a, p ,m, s are all complex operators
- 116. And for a very simplistic model where normalise complex operators, and combine processing and memory types we deduce that...
- 117. And for a very simplistic model where normalise complex operators, and combine processing and memory types we deduce that... Ic = I Ei - Eo I = I Ei - E[a.s ( 1 + m + p + mp ) ] I
- 118. And for a very simplistic model where normalise complex operators, and combine processing and memory types we deduce that... Ic = I Ei - Eo I = I Ei - E[a.s ( 1 + m + p + mp ) ] I Where: a, p ,m, s are all complex operators
- 119. After some mathematical juggling and normalisation, liberties, crimes, & of course engineering licence... Processor Memory ∝ log [1 + S. A (1 + P + M + P.M ) ] I 2 Sensor Actuator
- 120. And finally, something we can work with: Ic = k log2 [1 + K.S.A (1 + P + M + P.M ) ] System Constants For P.M >> (P + M) Ic ~ k log2 [1 + K.S.A (1 + P.M ) ] Relative Intelligence k log2 [1 + K.S.A (1 + P.M ) ] = of different systems Flops (or machine cycles)
- 121. Key Implication ! If I am correct: We are not racing toward a singularity as fast as people have projected
- 122. Key Implication ! If I am correct: We are not racing toward a singularity as fast as people have projected Increasing S. P. M. A by a factor of 1000 would only see intelligence grow 10 fold !
- 123. log2 Intelligence growth Intelligence 40 multiple 30 20 10 S. P. M. A 1 1 103 106 109 1012 1015 1018
- 124. log2 Intelligence growth Intelligence 40 multiple 30 20 10 S. P. M. A 1 1 103 106 109 1012 1015 1018
- 125. Remember - many systems are far far more complex than I have assumed Processor Memory Sensors 1 ShortTerm System Environs Processor Memory Actuator 2 LongTerm ...and this may defeat a full mathematical development & leave us with nothing more than a mere ‘rule of thumb’ understanding of the influencers of intelligence!
- 126. So why isn’t my laptop intelligent?
- 127. So why isn’t my laptop intelligent? How about: - sensors? - adaptability?
- 128. But we are rolling out intelligent sensors!
- 129. But we are rolling out intelligent sensors! And our machines will see, hear, smell…. …detect and communicate in a way and at a rate we cannot!
- 130. Life today! Processing Carbon = Silicon Memory Carbon = Silicon Comms Carbon < Silicon Sensors Carbon >> Silicon
- 131. What can we do with AI & AL now?
- 132. What can we do with AI & AL now?
- 134. Genetically ~99.4% human If ~0.6% matters so much… no wonder IT systems are so dumb!
- 135. In conclusion! Understanding non-linear systems presents us with our biggest & most important challenge & opportunity
- 136. In conclusion! Understanding non-linear systems presents us with our biggest & most important challenge & opportunity Whatever advance and advantage we have enjoyed through linear models, thinking, and endeavours, will be totally eclipsed by a move into the realm of non- linearity P Cochrane 1995
- 137. A world of great challenges and even greater opportunities…. Thank you, cochrane.org.uk ca-global.org COCHRANE a s s o c i a t e s