9 mesi per portare a compimento anche questo progetto: leggere quel capolavoro epistemologico-divulgativo che e’ Gödel, Escher, Bach: un’eterna ghirlanda brillante , che sentii nominare per la prima volta da Antonio the film-maker parecchi anni fa. Adesso ci vorrebbero altri 3 mesi per rileggerlo da capo e poter poi forse dire di averlo veramente capito, ma non penso che lo farò. Bei tempi quelli in cui leggevo 50 libri l’anno.
Gödel, Escher, Bach è un testo estremamente ricco di spunti, tant’è vero che ci sono almeno 6+1, e non 3, discipline che si intersecano: al problema dell’incompletezza intrinseca formal systems, as shown by Gödel's theorem, the possible paradoxes of self, illustrated by the strange rings in the designs of Escher, the characteristics of recursive functions, exemplified by the canons and fugues by Bach, followed by at least 3 other items. The paradoxical dialogues between Achilles, the Tortoise, the Crab and other characters in the style of Alice in Wonderland by Lewis Carroll, the research unit that includes all contradiction, typical of Zen, and means of transmission / alteration of the information gene studied by molecular biology. And all this is basically a prerequisite to the development of a theoretical paradigm useful to reflect on thought, conscience, representations, emotions, and the possibility of Artificial Intelligence. I do not know what I mean.
short, there's 2-3 that writing his doctoral thesis. Sin is only on vacation for ten days, and in addition to finally give the final blow to the text above, I also expected the study of networking and in particular a document on IPv6, also challenging in its own way ...
A nice thing but I can mention here. As almost everyone knows, my current job is to support some of Microsoft products. As I understand it, the current research in AI is not much more advanced than that considered by Hofstadter (who writes, I remember, in the 70) in that it has finally managed to program a supercomputer - Deep Blue - in order to beat the greatest chess masters, but from here to speak with intelligence comparable to that of humans - Hal 9000-style to be clear - there runs, but at the epistemological level we can make some interesting points.
The basic idea of \u200b\u200bHofstadter in this regard is that there is an isomorphism between the structure of the nervous system with its billions of neurons, synapses and electrical shock, and the basic structure of a computer, which we know is made of circuits and electrical states on / off, ie 0 and 1. The fundamental difference would be the degree of complexity ', ie the number of layers, from simple clusters of electric shock, and through the formation of signals which then are grouped into symbols of the upper level, separate the neural substrate of higher level entities such as thoughts, emotions, or self-consciousness. In essence, the impression that some features of the human being are of a quality 'different from what can never be reached by computer science, is only due to backwardness of our understanding of the mechanisms of formation of symbols, and then by' inability to program a machine to do so. The idea è veramente interessante, perché l’obiezione principale alla possibilità dell’intelligenza artificiale, consiste appunto nel dire che una macchina non potrà mai fare altro che ciò per cui è stata programmata. Tuttavia – risponde Hofstadter – questo è vero anche per gli essere umani: non esiste qualcosa – che si voglia chiamare anima, o intuizione, o altro – che sfugge alle possibilità insite al nostro substrato neuronale; è solo che, non essendo ancora venuti a capo che in minima parte alla complessità del rapporto fra simboli di livello superiore e substrato, ci SEMBRA di poter fare cose per cui “non siamo programmati”.
C’è anche da notare che questa impostazione è sicuramente meccanicista, ma NON E’ AFFATTO DETERMINISTA , perchè NON DICE che tutto sia già determinato al livello più basso. Dice certo che senza il substrato, nessun livello superiore sarebbe possibile, ma dice anche CHE OGNI NUOVO STRATO, OGNI LIVELLO DI SIMBOLI DAI PIU’ SEMPLICI AI PIU’ COMPLESSI, CREA SENSO: cioè, potremmo dire, rende pensabili cose che al livello precedente erano impensabili. Crea possibilità NUOVE. Dal riconoscimento di forme semplici, all’apprendere dalle proprie esperienze, all’introspezione, cioè a ciò che assomiglia sempre più alla nostra definizione intuitiva di intelligenza. Hofstadter fa l’esempio the anthill: the level of individual ants, it seems that the movement of each is random, or rather, is dictated by simple interactions nerve, but at the level of the anthill, the sense of a specialization of labor to the overall prosperity is apparent, and probably feed back on the interactions of lower level, helping to select those to be repeated and which to discard. The point is that the sense of shared prosperity can exist only at the highest level, that of the anthill, and not to the level of the individual ant.
What happens at the level of computers? We know that behind all of us is the binary system, the 0 and 1, the circuits. The question is: what kind of programs we would be able a produrre se avessimo sempre programmato soltanto in linguaggio macchina, cioè assegnando direttamente una serie di 0 e di 1 alla sequenza dei circuiti? Direi programmi estremamente poveri di senso, e non solo per il tempo enorme che sarebbe necessario a programmare funzioni estremamente semplici. Invece è stato inventato l’Assembler, cioè un LINGUAGGIO che raccoglie in singoli comandi un insieme di funzioni semplici, ma che a livello di codice macchina consistono in parecchi 0 e parecchi 1. L’atto di sussumere una serie di operazioni di base in un solo simbolo, dà la possibilità di mettere in relazione simboli di questo nuovo livello fra loro in un modo che al livello precedente NON RIUSCIVAMO A PENSARE . E’ esattamente how we create meaning: the new combinations can certainly be translated back into machine code - indeed should usually takes care of a software interpreter, because the computer "understands" only the machine code -, the fact remains that what has taken meaning to the second level, the first level would not have had. Obviously, the computer did not stop the Assembler, but has developed many high-level languages, each one more suitable to obtain certain results over others. What we say about Hofstadter is basically: we can say that all possibilities were GIA 'IN TRANSISTOR CIRCUITS AND CONTAINED IN THE BEGINNING? NO, the hardware is certainly the precondition of subsequent developments, but does not contain and is not determine them: the sense is produced by increasing the number of levels of symbolization. Simple example: we will not consider an intelligent machine that was able to answer questions posed by an operator, simply by checking the exact wording of the question in a huge database containing millions of possible questions, and returns the response associated with (and would happen anyway sooner or later the question is not in the database), consider it already a little bit smarter if the program contains logical and semantic analysis of structures capable of analyzing sentences and syntactic elements in fields such as noun, verb, adjective, e a campi semantici quali ‘la salute’ – ad esempio in domande quali “come stai?”. Domanda semplice? Quante generazioni di computer saranno necessarie prima che una macchina sia in grado di capirla?
Che c’entra tutto questo col supporto tecnico? Beh, nel supporto tecnico ripercorriamo i livelli al contrario. Si parte dal livello più alto, l’interfaccia grafica, e si ridiscende verso il codice; e si ridiscende di tanti livelli quanti necessari per risolvere lo specifico problema. A parte la ricerca di casi simili nel database interno, e l’eventuale esistenza di hotfix, si comincia dalla superficie. Sotto eventuali errori di configurazione correggibili già al livello più alto, may be necessary to use the command line (or shell, in Linux) to perform diagnostic tool of what happens at a lower level, or other proprietary tools that might have a new graphical interface, but see what happens "under "the surface, or enable logging, namely the recording formats text of what the program is doing at a certain level below. It is always in language, as Technical, read with some study and some practice. Even debugging, which requires significantly more expertise and is managed by technical escalation, and which aims to find a bug in the code was the basis of manifest error, be done at a higher level than that of the binary language code. But then the question may arise: why not everything is done programmatically? Or, why not, to prevent any problems in advance making sure that the code is bug-free? Why do I need this double movement from the bottom up in development (programming), and from complex to simple (binary) when troubleshooting? Each computer knows that it is impossible to write a program that is totally free of bugs: the limit of such a program is nothing, the zero lines of code. Every time you program, you increase the level of complexity whole, the greater the level of new information obtained, the more bugs are certainly introduced. More sense it makes more sense not (errors) is produced with it. And one of two things: either you decide to start from scratch each time virtually paralyzing the development of computer - any sense of equal minimum of risk - or consciously accept the challenge of complexity, built on the foundations of the previous level, recognizing the inevitability bugs are introduced, but also the enormous potential in terms of production of new information / direction. This approach, in addition to being deterministic, is not even Reductionism because if you translate the POST complex event from a more simple, it would be reduced to it, since it allows to derive, or deduct from the first second.
not understand why Hofstadter, after spending 700 pages to explain the isomorphism between the formal levels that enable us to understand Gödel's theorem, and levels that allow you to sketch a formal theory of mental processes, states online credit in principle to a full explanation of reductionist thinking, among other things, just a few lines after repeating that each level adds layers of meaning that simply does not exist at the lower level (in terms of 0 and 1, for example, that sense can you say "the program can no longer send mail?"). Or rather, it is clear that his main concern is to clear the field of mystical view of consciousness and free will. But exactly what does Gödel's theorem? He says: every sufficiently powerful formal system is incomplete in the sense that there is a true proposition within that system, but not certified (undecidable). The limit that Gödel puts the claims of absolute knowledge, is strong, but not absolute, in turn, it is plausible that the undecidable proposition in a system, it is perfectly sensible in a higher-level system, if able to definirlo. Così, Gödel mette la parola fine alla speranza dei Principia Mathematica di Russell e Whitehead , la speranza che la matematica, o qualsiasi altro sistema formale coerente, sia anche completo. Quindi i limiti sono sempre relativi a un livello/sistema specifico, ma sono anche inevitabili. Il punto cruciale è che il bagno di umiltà a cui Gödel costringe la matematica IMPONE DI ACCETTARE L’ESISTENZA DI LIMITI INTRINSECI, LIMITI ESSI STESSI FORMALI E NON DOVUTI ALL’INADEGUATEZZA DELLA SCIENZA DI FRONTE ALL’IMPERSCRUTABILITA’ DI DIO . Per quanto si diventi maestri nello spostare lo sguardo tra livelli, NON SI POSSONO OSSERVARE LIVELLI DIVERSI ALLO STESSO TEMPO, OGNI SGUARDO NE PREVENT ANOTHER, a bit like those drawings where you see different things depending which focuses attention on the figure or the background: you can never see the two together. Therefore, it makes little sense to say that you will understand the principles of intelligence when you will be able to master the jumps between levels through a concept of causality is flexible enough ...
is therefore technical support, as the levels that we must go down every time dependent from the specific case, is defined as the ATTIVITA'COMPLEMENTARE PRODUCTION CODE . Now, if we think about it, this pendulum movement, this form of feedback component on the other with a level on the other, the product manufacturer and vice versa, are known well before the advent of scientific epistemology and science of complexity theory. Just think of the relationship between theory and experiment in physics.
The new idea of \u200b\u200bArtificial Intelligence, is to see an isomorphism between the structure of the nervous system and circuits of computers, isomorphism legitimized by the universality of formal systems. Universality in Hofstadter also sometimes doubts. Because the basic problem is we are sure that the binary logic is the most appropriate? That our nervous system does not provide intermediate stages fra 0 e 1? Intanto si inizia a sentir parlare di computer quantistici...
La discussione sarebbe appena iniziata, ma ecco di nuovo la disgrazia: non è solo questione di tempo, ma anche il fatto che, mentre ho tutti i testi di informatica che mi servono direttamente in pdf, gli altri libri sono in versione cartacea, e non posso portarne con me più di 3-4 alla volta. E senza la possibilità di accedere continuamente alla mia preziosa biblioteca, vado poco lontano.
0 comments:
Post a Comment