July 2021 - nexa - server-nexa.polito.it

Re: [nexa] A Taxonomy of Judicial Technological Competence

by VETRO' ANTONIO

Rilevante che il primo principio sia (1) Being alert to technologically-based risks e non (1) Be proficient with techniques X,Y,Z Il giorno 26 lug 2021, alle ore 10:49, don Luca Peyron <dluca.universitari(a)gmail.com<mailto:dluca.universitari@gmail.com>> ha scritto: Interessanti sviluppi dal Canada Buona lettura dl http://www.slaw.ca/2021/06/24/a-taxonomy-of-judicial-technological-competen… _________________________ don Luca Peyron Direttore Ufficio per la Pastorale Universitaria Arcidiocesi di Torino www.universitari.to.it<http://www.universitari.to.it/> via XX settembre 83, Torino tel. 011 5156239 _______________________________________________ nexa mailing list nexa(a)server-nexa.polito.it<mailto:nexa@server-nexa.polito.it> https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa

July 26, 2021

An Ugly Truth

by Antonio Iacono

ZUCK: i have over 4000 emails, pictures, addresses, sns FRIEND: what!? how'd you manage that one? ZUCK: people just submitted it ZUCK: i don't know why ZUCK: they "trust me" ZUCK: dump fucks Quindici anni dopo i (...) che si "fidano" sono 2,80 miliardi qui https://www.mondadoristore.it/libri/Sheera-Frenkel/aut04009098/ il libro appena uscito di Sheera Frenkel e Cecilia Kang Antonio

July 26, 2021

A Taxonomy of Judicial Technological Competence

by don Luca Peyron

Interessanti sviluppi dal Canada Buona lettura dl http://www.slaw.ca/2021/06/24/a-taxonomy-of-judicial-technological-competen… _________________________ don Luca Peyron Direttore Ufficio per la Pastorale Universitaria Arcidiocesi di Torino www.universitari.to.it via XX settembre 83, Torino tel. 011 5156239

July 26, 2021

Re: [nexa] Is GitHub Copilot a blessing, or a curse? · fast.ai

by maurizio lana

per quanto vedo e conosco le medesime tecniche generative non producono gran risultati nemmeno nel linguaggio naturale. nel senso che i sistemi di IA sono conversatori da party eccellenti - quando uno, a un ricevimento di matrimonio, in un gruppo di sconosciuti, interloquisce con frasi gentili che raccolgono e riprendono quello che si sta dicendo. competenza sintattica ma non semantica, di quando si sa parlare ma non si sa che cosa dire. questo appare bene nell'articolo GPT-3. 2020. «A Robot Wrote This Entire Article. Are You Scared yet, Human?» /The Guardian/, 8 settembre 2020. http://www.theguardian.com/commentisfree/2020/sep/08/robot-wrote-this-artic… <http://www.theguardian.com/commentisfree/2020/sep/08/robot-wrote-this-artic…>. che in realtà è un mesh fatto da umani delle migliori parti di 8 diverse versioni dell'articolo che GPT-3 ha prodotto a partire da specifici seeds testuali forniti dagli umani stessi. cioè GPT-3 ha messo la sintassi, ma la semantica l'hanno costruita gli umani. e anche così la consequenzialità complessiva del (breve) discorso è sabbiosa. diverso, in parte, "1 the road" (Goodwin, Ross, Kenric McDowell, e Google. 2018. /1 the road/. Paris: JBE Books) dove a parte la configurazione del sistema di IA non c sono interventi umani. ma chi lo commenta come 'opera letteraria' non ne dice granché perché i singoli elementi di cui è fatto il libro mancano della tessitura complessiva, semantica. Maurizio Il 26/07/21 09:43, nexa-request(a)server-nexa.polito.it ha scritto: > Date: Mon, 26 Jul 2021 09:43:29 +0200 > From: Guido Vetere <vetere.guido(a)gmail.com> > To: maurizio lana <maurizio.lana(a)uniupo.it> > Cc: Nexa <nexa(a)server-nexa.polito.it> > Subject: Re: [nexa] Is GitHub Copilot a blessing, or a curse? · > fast.ai > Message-ID: > <CAD3hHB6ZLMqdtOEUzz7paQBFfbXLdBwxTsyje-VLCReOzCgTMQ(a)mail.gmail.com> > Content-Type: text/plain; charset="utf-8" > > no, non c'è pericolo. > è invece interessante osservare come le medesime tecniche generative > producano risultati soddisfacenti (almeno sotto qualche aspetto) se > applicate al linguaggio naturale, ma assolutamente casuali (29% di > accuratezza è nulla) se applicate al codice. il fatto è che per il codice > la semantica non è un 'optional' :-) > > G. ------------------------------------------------------------------------ il faut essayer, nous aussi. c'est ça, le progrès. à force d'essayer, peut-être qu'on aura à la fin les organes nécessaires, par exemple l'organe de la dignité, ou de la fraternité... " r. gary, les racines du ciel ------------------------------------------------------------------------ Maurizio Lana - 347 7370925

July 26, 2021

Memento | 90° Nexa Lunch Seminar | 28 Luglio 2021

by Nexa Media

Gentilissimi, vi ricordiamo che *mercoledì 28 ***luglio*, dalle ore 13.00 alle ore 14.00*, si svolgerà il 90° Nexa Lunch Seminar, con un incontro dal titolo: /*"Lo stato della censura Internet nel 2020: aggiornamenti dall'Osservatorio Aperto delle Interferenze di rete"*/. Ospite: *Arturo Filastò* (Fellow Centro Nexa). Sarà possibile partecipare al seminario connettendosi al seguente indirizzo: https://didattica.polito.it/VClass/NexaEvent <https://didattica.polito.it/VClass/NexaEvent> Maggiori informazioni sull'incontro sono disponibili alla pagina: https://nexa.polito.it/lunch-90 <https://nexa.polito.it/lunch-90> Cordiali saluti,   -- Anita Botta Communication Manager Nexa Center for Internet & Society Politecnico di Torino – DAUIN Corso Duca degli Abruzzi, 24 - 10129 Torino _web:_ https://nexa.polito.it/ <https://nexa.polito.it/> _mail:_anita.botta@polito.it <mailto:anita.botta@polito.it> tel: 011 090 7219

July 26, 2021

Re: [nexa] Is GitHub Copilot a blessing, or a curse? · fast.ai

by Guido Vetere

no, non c'è pericolo. è invece interessante osservare come le medesime tecniche generative producano risultati soddisfacenti (almeno sotto qualche aspetto) se applicate al linguaggio naturale, ma assolutamente casuali (29% di accuratezza è nulla) se applicate al codice. il fatto è che per il codice la semantica non è un 'optional' :-) G. On Mon, 26 Jul 2021 at 09:10, maurizio lana <maurizio.lana(a)uniupo.it> wrote: > in questo ambito è possibile che la moneta cattiva scacci quella buona? > cioè che il codice scritto dai sistemi di AI, anche se crappy, soppianti > quello buono /ottimo prodotto da analisti e programmatori capaci? > se è possibile, si può fare qualcosa cosa per evitarlo? > > nel senso che se già è così difficile [eufemismo] che sistemi di AI > prodotti da soggetti umani operino in modo eticamente e culturalmente > soddisfacente, c'è da chiedersi come possano operare dei crappy software > prodotti da sistemi di AI imperfetti [eufemismo] > Maurizio > > > > Il 25/07/21 12:00, nexa-request(a)server-nexa.polito.it ha scritto: > > Date: Sun, 25 Jul 2021 09:52:35 +0000 > > From: Alberto Cammozzo <ac+nexa(a)zeromx.net> > > To: Nexa <nexa(a)server-nexa.polito.it> > > Subject: [nexa] Is GitHub Copilot a blessing, or a curse? · fast.ai > > Message-ID: <E381DC52-434E-41A8-B178-D071057565EB(a)zeromx.net> > > Content-Type: text/plain; charset="utf-8" > > > > <https://www.fast.ai/2021/07/19/copilot> > > > > [...] > > According to OpenAI’s paper, Codex only gives the correct answer 29% of > the time. And, as we’ve seen, the code it writes is generally poorly > refactored and fails to take full advantage of existing solutions (even > when they’re in Python’s standard library). > > > > Copilot has read GitHub’s entire public code archive, consisting of tens > of millions of repositories, including code from many of the world’s best > programmers. Given this, why does Copilot write such crappy code? > > > > The reason is because of how language models work. They show how, on > average, most people write. They don’t have any sense of what’s correct or > what’s good. Most code on GitHub is (by software standards) pretty old, and > (by definition) written by average programmers. Copilot spits out it’s best > guess as to what those programmers might write if they were writing the > same file that you are. OpenAI discuss this in their Codex paper: > > > > “As with other large language models trained on a next-token > prediction objective, Codex will generate code that is as similar as > possible to its training distribution. One consequence of this is that such > models may do things that are unhelpful for the user” > > > > One important way that Copilot is worse than those average programmers > is that it doesn’t even try to compile the code or check that it works or > consider whether it actually does what the docs say it should do. Also, > Codex was not trained on code created in the last year or two, so it’s > entirely missing recent versions, libraries, and language features. For > instance, prompting it to create fastai code results only in proposals that > use the v1 API, rather than v2, which was released around a year ago. > > > > Complaining about the quality of the code written by Copilot feels a bit > like coming across a talking dog, and complaining about its diction. The > fact that it’s talking at all is impressive enough! > > > > Let’s be clear: The fact that Copilot (and Codex) writes > reasonable-looking code is an amazing achievement. From a machine learning > and language synthesis research point of view, it’s a big step forward. > > > > But we also need to be clear that reasonable-looking code that doesn’t > work, doesn’t check edge cases, and uses obsolete methods, and is verbose > and creates technical debt, can be a big problem. > > > > > > > ------------------------------------------------------------------------ > > the knowledge gap between rich and poor is widening > witten, bainbridge, nichols, how to build a digital library > > ------------------------------------------------------------------------ > Maurizio Lana - 347 7370925 > > _______________________________________________ > nexa mailing list > nexa(a)server-nexa.polito.it > https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa >

July 26, 2021

Re: [nexa] Is GitHub Copilot a blessing, or a curse? · fast.ai

by Alberto Cammozzo

Evolutivamente il codice dominante non è quello migliore, ma quello che si riproduce di più. La strategia riproduttiva del codice di copilot è quella di facilitare il compito al programmatore, non di produrre codice migliore. Teoricamente, se copilot venisse adottato su larga scala, il suo codice dovrebbe diventare dominante. Dovrà prima o poi vedersela con competizione di altri strumenti dello stesso tipo. Ciao, Alberto On July 26, 2021 7:10:35 AM UTC, maurizio lana <maurizio.lana(a)uniupo.it> wrote: >in questo ambito è possibile che la moneta cattiva scacci quella buona? >cioè che il codice scritto dai sistemi di AI, anche se crappy, >soppianti >quello buono /ottimo prodotto da analisti e programmatori capaci? >se è possibile, si può fare qualcosa cosa per evitarlo? > >nel senso che se già è così difficile [eufemismo] che sistemi di AI >prodotti da soggetti umani operino in modo eticamente e culturalmente >soddisfacente, c'è da chiedersi come possano operare dei crappy >software >prodotti da sistemi di AI imperfetti [eufemismo] >Maurizio > > > >Il 25/07/21 12:00, nexa-request(a)server-nexa.polito.it ha scritto: >> Date: Sun, 25 Jul 2021 09:52:35 +0000 >> From: Alberto Cammozzo <ac+nexa(a)zeromx.net> >> To: Nexa <nexa(a)server-nexa.polito.it> >> Subject: [nexa] Is GitHub Copilot a blessing, or a curse? · fast.ai >> Message-ID: <E381DC52-434E-41A8-B178-D071057565EB(a)zeromx.net> >> Content-Type: text/plain; charset="utf-8" >> >> <https://www.fast.ai/2021/07/19/copilot> >> >> [...] >> According to OpenAI’s paper, Codex only gives the correct answer 29% >of the time. And, as we’ve seen, the code it writes is generally poorly >refactored and fails to take full advantage of existing solutions (even >when they’re in Python’s standard library). >> >> Copilot has read GitHub’s entire public code archive, consisting of >tens of millions of repositories, including code from many of the >world’s best programmers. Given this, why does Copilot write such >crappy code? >> >> The reason is because of how language models work. They show how, on >average, most people write. They don’t have any sense of what’s correct >or what’s good. Most code on GitHub is (by software standards) pretty >old, and (by definition) written by average programmers. Copilot spits >out it’s best guess as to what those programmers might write if they >were writing the same file that you are. OpenAI discuss this in their >Codex paper: >> >> “As with other large language models trained on a next-token >prediction objective, Codex will generate code that is as similar as >possible to its training distribution. One consequence of this is that >such models may do things that are unhelpful for the user” >> >> One important way that Copilot is worse than those average >programmers is that it doesn’t even try to compile the code or check >that it works or consider whether it actually does what the docs say it >should do. Also, Codex was not trained on code created in the last year >or two, so it’s entirely missing recent versions, libraries, and >language features. For instance, prompting it to create fastai code >results only in proposals that use the v1 API, rather than v2, which >was released around a year ago. >> >> Complaining about the quality of the code written by Copilot feels a >bit like coming across a talking dog, and complaining about its >diction. The fact that it’s talking at all is impressive enough! >> >> Let’s be clear: The fact that Copilot (and Codex) writes >reasonable-looking code is an amazing achievement. From a machine >learning and language synthesis research point of view, it’s a big step >forward. >> >> But we also need to be clear that reasonable-looking code that >doesn’t work, doesn’t check edge cases, and uses obsolete methods, and >is verbose and creates technical debt, can be a big problem. >> >> > > >------------------------------------------------------------------------ > >the knowledge gap between rich and poor is widening >witten, bainbridge, nichols, how to build a digital library > >------------------------------------------------------------------------ >Maurizio Lana - 347 7370925 > >_______________________________________________ >nexa mailing list >nexa(a)server-nexa.polito.it >https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa

July 26, 2021

Is GitHub Copilot a blessing, or a curse? · fast.ai

by maurizio lana

in questo ambito è possibile che la moneta cattiva scacci quella buona? cioè che il codice scritto dai sistemi di AI, anche se crappy, soppianti quello buono /ottimo prodotto da analisti e programmatori capaci? se è possibile, si può fare qualcosa cosa per evitarlo? nel senso che se già è così difficile [eufemismo] che sistemi di AI prodotti da soggetti umani operino in modo eticamente e culturalmente soddisfacente, c'è da chiedersi come possano operare dei crappy software prodotti da sistemi di AI imperfetti [eufemismo] Maurizio Il 25/07/21 12:00, nexa-request(a)server-nexa.polito.it ha scritto: > Date: Sun, 25 Jul 2021 09:52:35 +0000 > From: Alberto Cammozzo <ac+nexa(a)zeromx.net> > To: Nexa <nexa(a)server-nexa.polito.it> > Subject: [nexa] Is GitHub Copilot a blessing, or a curse? · fast.ai > Message-ID: <E381DC52-434E-41A8-B178-D071057565EB(a)zeromx.net> > Content-Type: text/plain; charset="utf-8" > > <https://www.fast.ai/2021/07/19/copilot> > > [...] > According to OpenAI’s paper, Codex only gives the correct answer 29% of the time. And, as we’ve seen, the code it writes is generally poorly refactored and fails to take full advantage of existing solutions (even when they’re in Python’s standard library). > > Copilot has read GitHub’s entire public code archive, consisting of tens of millions of repositories, including code from many of the world’s best programmers. Given this, why does Copilot write such crappy code? > > The reason is because of how language models work. They show how, on average, most people write. They don’t have any sense of what’s correct or what’s good. Most code on GitHub is (by software standards) pretty old, and (by definition) written by average programmers. Copilot spits out it’s best guess as to what those programmers might write if they were writing the same file that you are. OpenAI discuss this in their Codex paper: > > “As with other large language models trained on a next-token prediction objective, Codex will generate code that is as similar as possible to its training distribution. One consequence of this is that such models may do things that are unhelpful for the user” > > One important way that Copilot is worse than those average programmers is that it doesn’t even try to compile the code or check that it works or consider whether it actually does what the docs say it should do. Also, Codex was not trained on code created in the last year or two, so it’s entirely missing recent versions, libraries, and language features. For instance, prompting it to create fastai code results only in proposals that use the v1 API, rather than v2, which was released around a year ago. > > Complaining about the quality of the code written by Copilot feels a bit like coming across a talking dog, and complaining about its diction. The fact that it’s talking at all is impressive enough! > > Let’s be clear: The fact that Copilot (and Codex) writes reasonable-looking code is an amazing achievement. From a machine learning and language synthesis research point of view, it’s a big step forward. > > But we also need to be clear that reasonable-looking code that doesn’t work, doesn’t check edge cases, and uses obsolete methods, and is verbose and creates technical debt, can be a big problem. > > ------------------------------------------------------------------------ the knowledge gap between rich and poor is widening witten, bainbridge, nichols, how to build a digital library ------------------------------------------------------------------------ Maurizio Lana - 347 7370925

July 26, 2021

The FTC Votes Unanimously to Enforce Right to Repair | WIRED

by Alberto Cammozzo

<https://www.wired.com/story/ftc-votes-to-enforce-right-to-repair/> [...] The FTC’s endorsement of the rules is not a surprise outcome; the issue of Right to Repair has been a remarkably bipartisan one, and the FTC itself issued a lengthy report in May that blasted manufacturers for restricting repairs. But the 5 to 0 vote signals the commission’s commitment to enforce both federal antitrust laws and a key law around consumer warranties—the Magnuson Moss Warranty Act—when it comes to personal device repairs. The vote, which was led by new FTC chair and known tech critic Lina Khan, also comes 12 days after President Joe Biden signed a broad executive order aimed at promoting competition in the US economy. The order addressed a wide range of industries, from banks to airlines to tech companies. But a portion of it encouraged the FTC, which operates as an independent agency, to create new rules that would prevent companies from restricting repair options for consumers. “When you buy an expensive product, whether it's a half-a-million-dollar tractor or a thousand-dollar phone, you are in a very real sense under the power of the manufacturer,” says Tim Wu, special assistant to the president for technology and competition policy within the National Economic Council. “And when they have repair specifications that are unreasonable, there's not a lot you can do." Wu added that Right to Repair has become a "visceral example" of the enormous imbalance between workers, consumers, small businesses, and larger entities. [...]

July 25, 2021

Is GitHub Copilot a blessing, or a curse? · fast.ai

by Alberto Cammozzo

<https://www.fast.ai/2021/07/19/copilot> [...] According to OpenAI’s paper, Codex only gives the correct answer 29% of the time. And, as we’ve seen, the code it writes is generally poorly refactored and fails to take full advantage of existing solutions (even when they’re in Python’s standard library). Copilot has read GitHub’s entire public code archive, consisting of tens of millions of repositories, including code from many of the world’s best programmers. Given this, why does Copilot write such crappy code? The reason is because of how language models work. They show how, on average, most people write. They don’t have any sense of what’s correct or what’s good. Most code on GitHub is (by software standards) pretty old, and (by definition) written by average programmers. Copilot spits out it’s best guess as to what those programmers might write if they were writing the same file that you are. OpenAI discuss this in their Codex paper: “As with other large language models trained on a next-token prediction objective, Codex will generate code that is as similar as possible to its training distribution. One consequence of this is that such models may do things that are unhelpful for the user” One important way that Copilot is worse than those average programmers is that it doesn’t even try to compile the code or check that it works or consider whether it actually does what the docs say it should do. Also, Codex was not trained on code created in the last year or two, so it’s entirely missing recent versions, libraries, and language features. For instance, prompting it to create fastai code results only in proposals that use the v1 API, rather than v2, which was released around a year ago. Complaining about the quality of the code written by Copilot feels a bit like coming across a talking dog, and complaining about its diction. The fact that it’s talking at all is impressive enough! Let’s be clear: The fact that Copilot (and Codex) writes reasonable-looking code is an amazing achievement. From a machine learning and language synthesis research point of view, it’s a big step forward. But we also need to be clear that reasonable-looking code that doesn’t work, doesn’t check edge cases, and uses obsolete methods, and is verbose and creates technical debt, can be a big problem. [...]

July 25, 2021