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 <[email protected]> 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 <[email protected]> per conto di maurizio lana < > [email protected]> > 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 > > > >
_______________________________________________ nexa mailing list [email protected] https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa
