300 likes | 452 Views
COGNITIVE SCIENCE 2 KOGNITIVNA ZNANOST 2. M. Gams Institut „ Jo ž ef Stefan “. FIRST ANALYSIS. Easy – Hard question - Easy – how to make AI? - Hard – to explain consciousness, why and how it appeared ... Achievements in the computer age?
E N D
COGNITIVE SCIENCE 2 KOGNITIVNA ZNANOST 2 M. Gams Institut „Jožef Stefan“
FIRST ANALYSIS • Easy – Hard question • - Easy – how to make AI? • - Hard – to explain consciousness, why and how it appeared ... • Achievements in the computer age? robots (work, walk …)engineering intelligence (chess, applications)runterstanding (Turing test) Fast technol. progress, slow cognitive
Artificial intelligence – how to make computers intelligent, Cognitive science - human-like computers
Brain capacity • 500 generations Europe • 5000 generations pra-Eva • Die-out
Where is the smart computer? Unknown barrier
PARADOXES • Empirical lack of true human-level AI • Sloman’s paradox – Einstein’s book • Searl’s paradox Chinese room • Chalmers zombie consciousnessThe notion of a philosophical zombie is used mainly in thought experiments intended to support arguments (often called "zombie arguments") against forms of physicalism such as materialism, behaviorism and functionalism. Physicalism is the idea that all aspects of human nature can be explained by physical means: specifically, all aspects of human nature and perception can be explained from a neurobiological standpoint.
Roger Penrose • Oxford, črne luknje, Hawkings … • Controversial books on the connection between fundamental physics and human consciousness. In The Emperor's New Mind (1989), he argues that known laws of physics are inadequate to explain the phenomenon of human consciousness. Penrose hints at the characteristics this new physics may have and specifies the requirements for a bridge between classical and quantum mechanics (what he terms correct quantum gravity, CQG). He claims that the present computer is unable to have intelligence because it is a deterministic system that for the most part simply executes algorithms, as a billiard table where billiard balls act as message carriers and their interactions act as logical decisions. He argues against the viewpoint that the rational processes of the human mind are completely algorithmic and can thus be duplicated by a sufficiently complex computer -- this is in contrast to views, e.g., biological naturalism, that human behavior but not consciousness might be simulated. This is based on claims that human consciousness transcends formal logic systems because things such as the insolubility of the halting problem and Gödel's incompleteness theorem restrict an algorithmically based logic from traits such as mathematical insight.
Roger Penrose In 1994, Penrose followed up The Emperor's New Mind with Shadows of the Mind and in 1997 with The Large, the Small and the Human Mind, further updating and expanding his theories. Penrose's views on the human thought process are not widely accepted in scientific circles. According to Marvin Minsky, because people can construe false ideas to be factual, the process of thinking is not limited to formal logic. Furthermore, he says that artificial intelligence (AI) programs can also conclude that false statements are true, so error is not unique to humans. Penrose and Stuart Hameroff have speculated that human consciousness is the result of quantum gravity effects in microtubules.
Quantum physics • interpretation of the Schrëdinger equation: • - classical or Copenhagen or the GRW interpretation • - Bohm's interpretation • - multiple-worlds interpretation.
Quantum computing • Deutch • Theory well defined, practical prototypes • In principle not more powerful than Turing machines • Can perform differently
Supercomputing Turing machine with oracle – TuringSpecific quantum computing – KomarTrial-and-error – PutnamExtended Turing machines – with real numbers – AbramsonMcCulloch-Pitts neurons – growing - Karp and Lipton Analog – non simulatable - Rubel, KononenkoMultiple computing - Gams Interaction machines – WegnerCoupled TM – open input - Copeland, SylvanPartially random machines – truly random – Copeland, Turing
ZEUS MACHINE Zeus machine - Boolos and Jeffrey 1974 – infinite computing, each step is computed faster and faster Example – computing/going from A to B, first half in 1sec, next quarter in ¼ … end in 2 sec thus computing infinite numbers The Supermind book, subtitled People harness hypercomputation and more, authored by Selmer Bringsjord and Michael Zenzen, aggressively attacks the strong-AI viewpoint that human thinking processes are computationally as strong as computers Bringsjord, S. and Zenzen, M. J. (2003), Superminds, Kluwer.
M. Gams: Weak intelligence: Through the principle and paradox of multiple knowledge, Advances in computation: Theory and practice, Volume 6, Nova science publishers, inc., NY, ISBN 1-56072-898-1, pp. 245, 2001. • Najboljše rezultate je možno dosegati le ob uporabi mnogoterih modelov (kibernetično). • Miselni procesi so mnogoteri. Povečana računska/miselna sposobnost prihaja iz mnogoterih procesov, ki interaktirajo med seboj (teoretično, inteligentno). V principu je ta računski mehanizem močnejši kot univerzalni digitalni računalnik oz. Turingov stroj. Predstavitev osnovne teze – principa mnogoterosti (1985-2001)
Formalni/matematični (od 2 do 10 samostojnih modelov); ob predpostavkah realnega sveta pričakovani boljši rezultati; smiselno kombiniranje-integriranje ob razumnih predpostavkah (bolje kot 50%) • Simulacije modelov z različnimi metodami in parametri kažejo podobno • Teoretične analize (Turingovi stroji) • Študij ljudi (mnogoterosti možganov sedaj in v preteklosti; skupine ljudi) • Empirične meritve sistemov • Podobnost s fiziko (Heisenberg, teorija večih svetov) Potrditve osnovne teze
Predstavitev osnovne teze – principa mnogoterosti (1985-2001)Turingov stroj
Predstavitev osnovne teze – principa mnogoterosti (1985-2001) Wegner 1997 – interakcija močnejša
Ljudje • Delamo najbolje v skupinah (več glav več ve; slabo – preveč kuharjev, slaba juha) • Človek + računalnik bolje kot samo človek ali samo računalnik • Študij možganov – dve hemisferi; razvoj človeških možganov – čedalje bolj mnogoteri, študij opic • Študij možganov – corpus calosum, split-brain research, moški-ženski, dve hemisferi – izrazito mnogoteri • Potrjujejo tezo o principu mnogoterosti pri ljudeh
Empirične potrditve - kibernetika, umetna inteligenca, strojno učenje - boljša klasifikacijska točnost - empirično: na tisoče meritev-potrditev- možno je preveriti model na konkretni aplikaciji s prilagoditvijo parametrov modela- več podobnih ugotovitev na specifičnih področjih (statistika, prepoznavanje vzorcev …)- omogočena ocena algoritmov vnaprej- omogočena analiza delovanja algoritmov (intuitivno in formalistično)- omogočeno snovanje boljših algoritmov
Multiple-worlds/quantum computing • Travel in space (back, forward, but one life) • How many universes, where?- physical (more dimensions)- mental- potential in future • Quantum computing – drugačno računanje, primitivni prototipi /40
Analogija s fiziko – paradoksi, dograditev znanstvenih teorij • Kvantna fizika, teorija večih svetov, najbolj široka izmed interpretacij (premočna, kje je neskončno svetov – v glavi, mentalno, ali fizično, potovanja v času??), resna znanstvena teorija, dr. Pavšič • Ali niso te teorije preveč sofisticirane? Tako fizikalne, kvantne kot mentalne? Recimo – kje v glavi je neskončno svetov, kje je veliko osebnosti (miselnih procesov), zakaj internet ni inteligenten?, zakaj agenti niso inteligentni?, ali znamo narediti mnogotere inteligentne računalnike? • Precej odprtih vprašanj, nejasnosti, vendar znanstvene teorije držijo- primerjajmo z drugimi principi. Heisenbergov princip, teorija večih svetov
Podobno kot Heisenbergov princip ločimo med sedanjimi in pravimi inteligentnimi sistemi. • Šibka inteligenca – Zakaj računalniki ne bodo nikoli mislili (razen če ne bodo drugače narejeni)?(namesto enega računalnika skoraj zadošča internet) • Za doseganje dobrih rezultatov nujne mnogotere metode • Paradoks mnogoterega znanja: več modelov = en model? statično - dinamično Posledice osnovne teze
PREFACE 1 ARTIFICIAL INTELLIGENCE 1.1 Artificial Intelligence Directions 1.2 History of Artificial Intelligence 1.3 Where's the AI? 1.4 Storage/Memory vs. Processing/Thinking 1.5 Problems with Formalistic AI 1.6 Strong-AI Super-Projects 2 TRENDS OF COMPUTER PROGRESS 3 THE BRAIN 4 STRONG VERSUS WEAK AI 4.1 Description 4.2 Sloman's Engineering Gradation of Strong-Weak AI 5 FUNDAMENTALS OF AI, COMPUTER SCIENCE AND SCIENCE IN GENERAL 5.1 Alan Turing 5.2 The Turing Test 5.3 Turing Machine and Church-Turing Thesis 5.4 Church-Turing Thesis and Turing Machines 5.5 Goedel's Theorem and the Halting Problem 5.6 Penrose's Analyses of Goedel's Theorem 5.7 Is Interaction Stronger than Algorithms? 6 THE PRINCIPLE AND PARADOX OF MULTIPLE KNOWLEDGE 6.1 Basic Definitions 6.2 The Principle of Multiple Knowledge 6.3 The Paradox of Multiple Knowledge 7 CONFIRMATIONS OF THE PRINCIPLE 7.1 Multiple Knowledge in Empirical Learning 7.2 Simulated Multiple Models 7.3 Formal Worst-Case Analyses 7.4 Formal Average-Case Improvements 7.5 Fitting the Model to Real-Life Applications 7.6 Human Multiple Reasoning 7.7 Cognitive Sciences and Common Sense Weak intelligence through the principle and paradox of multiple knowledge
8 CONSEQUENCES 8.1 Occam's Razor Vs. Multiple Knowledge 8.2 Bayes' Classifier And Multiple Knowledge 8.3 Properties of Knowledge 9 MANY-WORLDS THEORY AND QUANTUM COMPUTING 9.1 Paradoxes of Modern Physics 9.2 Interpretations of Quantum Physics 9.3 The Many-Worlds Theory 9.4 Objections to the Many-Worlds Interpretation 9.4 Quantum Computing 9.5 From Many Worlds to the Principle of Multiple Knowledge 10 STRONG AI FIGHTS BACK 11 CONCLUSION Weak intelligence through the principle and paradox of multiple knowledge
Povečano razumevanje področja inteligence in zavesti, intenziviranje raziskav v smeri umetne inteligence, bistveno povečane možnosti novih odkritij • Popravki obstoječih osnovnih teorij – Occamovega rezila, Church-Turingove teze, Turingovega stroja • Princip mnogoterosti je osnovni znanstveni princip Posledice