Siamo davvero irripetibili? Mi appassiona la domanda: davvero i nostri cervelli funzionano diversamente dall'IA? Cordialmente, Duccio (Alessandro Marzocchi) Il 9 feb 2024 alle ore 12:00 <nexa-request@server-nexa.polito.it> Daniela Tafani ha scritto:
Grazie della segnalazione, Maurizio. Sono domande che Joseph Weizenbaum poneva, anche a proposito della comprensione del linguaggio naturale, in Computer power and human reason:
The second kind of computer application that ought to be avoided, or at least not undertaken without very careful forethought, is that which can easily be seen to have irreversible and not entirely foreseeable side effects. If, in addition, such an application cannot be shown to meet a pressing human need that cannot readily be met in any other way, then it ought not to be pursued. The latter stricture follows directly from the argument I have already presented about the scarcity of human intelligence. The example I wish to cite here is that of the automatic recognition of human speech. There are now three or four major projects in the United States devoted to enabling computers to understand human speech, that is, to programming them in such a way that verbal speech directed at them can be converted into the same internal representations that would result if what had been said to them had been typed into their consoles. The problem, as can readily be seen, is very much more complicated than that of natural-language understanding as such, for in order to understand a stream of coherent speech, the language in which that speech is rendered must be understood in the first place. The solution of the "speech-understanding problem" therefore presupposes the solution of the "natural-language-understanding problem." And we have seen that, for the latter, we have only "the tiniest bit of relevant knowledge." But I am not here concerned with the technical feasibility of the task, nor with any estimate of just how little or greatly optimistic we might be about its completion. Why should we want to undertake this task at all? I have asked this question of many enthusiasts for the project. The most cheerful answer I have been able to get is that it will help physicians record their medical notes and then translate these notes into action more efficiently. Of course, anything that has any ostensible connection to medicine is automatically considered good. But here we have to remember that the problem is so enormous that only the largest possible computers will ever be able to manage it. In other words, even if the desired system were successfully designed, it would probably require a computer so large and therefore so expensive that only the largest and best-endowed hospitals could possibly afford it—but in fact the whole system might be so prohibitively expensive that even they could not afford it. The question then becomes, is this really what medicine needs most at this time? Would not the talent, not to mention the money and the resources it represents, be better spent on projects that attack more urgent and more fundamental problems of health care? But then, this alleged justification of speech-recognition "research" is merely a rationalization anyway. (I put the word "research" in quotation marks because the work I am here discussing is mere tinkering. I have no objection to serious scientists studying the psycho-physiology of human speech recognition.) If one asks such questions of the principal sponsor of this work, the Advanced Research Projects Agency (ARPA) of the United States Department of Defense, as was recently done at an open meeting, the answer given is that the Navy hopes to control its ships, and the other services their weapons, by voice commands. This project then represents, in the eyes of its chief sponsor, a long step toward a fully automated battlefield. I see no reason to advise my students to lend their talents to that aim. I have urged my students and colleagues to ask still another question about this project: Granted that a speech-recognition machine is bound to be enormously expensive, and that only governments and possibly a very few very large corporations will therefore be able to afford it, what will they use it for? What can it possibly be used for? There is no question in my mind that there is no pressing human problem that will more easily be solved because such machines exist. But such listening machines, could they be made, will make monitoring of voice communication very much easier than it now is. Perhaps the only reason that there is very little government surveillance of telephone conversations in many countries of the world is that such surveillance takes so much manpower. Each conversation on a tapped phone must eventually be listened to by a human agent. But speech-recognizing machines could delete all "uninteresting" conversations and present transcripts of only the remaining ones to their masters. I do not for a moment believe that we will achieve this capability within the future so clearly visible to Newell and Simon. But I do ask, why should a talented computer technologist lend his support to such a project? As a citizen I ask, why should my government spend approximately 2.5 million dollars a year (as it now does) on this project?
https://archive.org/details/computerpowerhum0000weiz_v0i3/page/270
Un saluto, Daniela ________________________________________ Da: nexa <nexa-bounces@server-nexa.polito.it> per conto di maurizio lana < maurizio.lana@uniupo.it> Inviato: giovedì 8 febbraio 2024 16:57 A: NEXA ML Oggetto: [nexa] sistemi di IA e scrittura
mi sono imbattuto in questo libro: Baron, Naomi S. Who wrote this? how AI and the lure of efficiency threaten human writing. Stanford, California: Stanford University Press, 2023.
nel capitolo finale l'autrice scrive: Imagine a world where AI’s current writing challenges have been solved. Where large language models (or their successors) don’t churn out ugliness. Where using them is energy efficient. Where predictive texting, spellcheck, and grammar programs are infallible. Where AI can produce lengthy texts that are non-repetitive, stylistically interesting, factually accurate, and always on topic. Oh, and can generate text that’s indistinguishable from what you might have written. Where would this world leave us humans? As we weigh options, keep in mind potential blowback of getting what we wish for. Cultural lore—be it of King Midas in Greek mythology, the recurrent “three wishes” stories across European tales, or W. W. Jacobs’s more modern “The Monkey’s Paw”—reminds us that attractive prospects may bear unforeseen consequences. lo trovo interessante perché molta parte della riflessione critica sui sistemi di AI si appunta su singoli aspetti mal-funzionanti/dis-funzionanti. mentre qui c'è una riflessione critica globale, 'a prescindere', che si esprime sui sistemi di IA ma che riguarda ogni ambito: è desiderabile/quali conseguenze ha, che in ogni campo ogni necessità sia risolta e ogni difficoltà operativa sia eliminata ? "teletrasporto per tutti a costo zero" e via in cima al monte Bianco con un tasto (in questo esempio si vede che sono proprio uno al pie' dei monti - piemontese): che cosa mi significherebbe?
Maurizio
________________________________
quanti nella loro vita si fecero custodi delle termopili, sono degni di più grande onore se prevedono (e molti lo prevedono) che all’ultimo comparirà un efialte e comunque i persiani passeranno kostantinos kavafis, termopili
________________________________ Maurizio Lana Università del Piemonte Orientale Dipartimento di Studi Umanistici Piazza Roma 36 - 13100 Vercelli