August 2024 - nexa - server-nexa.polito.it

Re: [nexa] “One day they might come for you”

by Roberto Resoli

Sulla stessa linea è di recente uscito "Fake You – An Activist’s Guide to Defeating Disinformation" di Simona Levi et al. https://xnet-x.net/en/fakeyou-disinformation-free-download/ > Conferences about disinformation are mainstream. They are appealing, and institutions love them. They all seem to follow the same formula: a star speaker, boasting a fashionable biography (that omits the financial or client-affective ties to a political party) rattles off a list of stereotypical evils of technology, leading to a conclusion that could be summarised like this: “Given the very new danger of disinformation and fake news brought by the Internet and Artificial Intelligence (Al), for your own sake, the solution is to create institutions that ensure that internet and digital are not evil and - basically - ‘regulate’ freedom of expression in the digital era for the sake of the Truth.” > > This false conclusion is the main reason for this book: fake news is used as an excuse to curtail civil rights. rob Il 10/08/24 16:12, J.C. DE MARTIN ha scritto: > *“One day they might come for you” > */Digital rights activist Andrew Lowenthal on progressives' support for > the censorship-industrial complex > / > Thomas Fazi > > Aug 05, 2024 > > In this guest post, Maike Gosch, a German communication strategist, > writer and former lawyer — whose article on the banning of the German > magazine Compact I published recently — interviews Andrew Lowenthal, the > founder and managing director of the digital civil liberties > organisation liber-net. > > Lowenthal is an Australian digital rights activist of German-Jewish > descent. For almost 18 years he was the Executive Director of > EngageMedia, an Asia-based NGO focused on human rights online, freedom > of expression and open technology. The digital rights environment in > which Lowenthal spent most of his adult life was avowedly progressive — > as was Lowenthal himself. > > [...] > > continua qui: https://www.thomasfazi.com/p/one-day-they-might-come-for-you > >

Aug. 12, 2024

Re: [nexa] “One day they might come for you”

by J.C. DE MARTIN

Restando a un livello molto generale, premesso che: - in anni recenti (post 2016, a spanne) il termine "populista" è stato volutamente distorto per essere usato per squalificare a priori posizioni sgradite, evitando di entrare nel merito; - da molto tempo, ma anche in questo caso con una recrudescenza in questi ultimi anni, il termine "complottista" è usato per squalificare a priori posizioni sgradite, evitando di entrare nel merito; sono convinto che ci siano forti motivi per essere preoccupati per il futuro della libertà di espressione anche nelle nostre democrazie liberali. Argomento che va discusso - come tutti gli altri - facendo attenzione solo al merito delle questioni, evitando etichettature e altre fallacie argomentative. E credo che, restando nel merito, ormai ci siano parecchi fatti, iniziative e dichiarazioni a giustificazione della mia preoccupazione. juan carlos On 12/08/24 11:32, 380° via nexa wrote: > Buongiorno, > > felice di vedere che, seppur faticosamente, pian piano "conspiracy is > the new normal", specialmente quando è di stampo POPULISTA :-D > > perché inutile girarci in giro: Lowenthal è l'archetipo del complottista > (a sua insaputa?), con decisi tratti populisti... o no?!? > > "J.C. DE MARTIN" <juancarlos.demartin(a)polito.it> writes: > >> *“One day they might come for you” >> */Digital rights activist Andrew Lowenthal on progressives' support for >> the censorship-industrial complex >> / >> Thomas Fazi >> >> Aug 05, 2024 >> >> In this guest post, Maike Gosch, a German communication strategist, >> writer and former lawyer — whose article on the banning of the German >> magazine Compact I published recently — interviews Andrew Lowenthal, the >> founder and managing director of the digital civil liberties >> organisation liber-net. >> >> Lowenthal is an Australian digital rights activist of German-Jewish >> descent. For almost 18 years he was the Executive Director of >> EngageMedia, an Asia-based NGO focused on human rights online, freedom >> of expression and open technology. The digital rights environment in >> which Lowenthal spent most of his adult life was avowedly progressive — >> as was Lowenthal himself. >> continua qui: https://www.thomasfazi.com/p/one-day-they-might-come-for-you > siccome non vorrei che i lettori si perdessero il nocciolo della > questione, ovvero la metamorfosi del progressismo di sinistra, > l'articolo continua così: > > --8<---------------cut here---------------start------------->8--- > > Then came the pandemic, and Lowenthal, as many of us, experienced an > anthological shock that led him to reconsider much of his worldview — > especially about his activist milieu and the progressive left more in > general. > > [...] Increasingly estranged from the progressive > former-digital-rights-now-turned-pro-censorship movement, Lowenthal > began to focus his work on the emerging digital censorship regime — or > censorship-industrial complex — even collaborating with journalists like > Matt Taibi and Michael Shellenberger on the Twitter Files. This > eventually led to the founding of liber-net. > > To stay on top of all things online censorship-related, make sure to > follow Lowenthal’s great Substack, Network Affects [1]. > > Lowenthal: [...] got very interested in left-wing activism [...] and > particularly what was known was “anti-globalisation”. [...] I helped > organise, in Melbourne against the World Economic Forum [...] then I > started liber-net, which focuses on digital civil liberties and digital > authoritarianism in the West, issues including censorship, free speech, > surveillance, ways in which technologies increasingly create digital > gates and barriers around access, such as what we saw during Covid with > the vaccine passports, that are now coming more, in terms of digital ID > issues around programmable currency, so essentially systems of digital > social control. > > Lowenthal: [...] (sulla Westminster Declaration [2], n.d.r.) the media > on the other hand rather dismissed it, or they just focus on the > right-wing people who signed it and ignored the left-wing people who > signed it to try and suggest that it was very bad, and you should stay > away from it. > > MG: Often the reaction in Germany when people hear about the > censorship-industrial complex is that it gets called a conspiracy > theory. What would your reply be to that? What do you think of that > criticism? > > Lowenthal: There’s huge amount of evidence now that there are a massive > number of organisations coordinating to censor content on the > internet. In fact, a lot of liberal and progressive groups now > acknowledge that this is happening. At the start of the Twitter Files, > the line was, “Oh, no, it’s a conspiracy theory. None of this is > happening, etc.”. Now, most people are saying: “Well, yes, the > government and NGOs are involved in content moderation. But it’s > okay. It’s legal. Governments should be paying attention and protecting > people from misinformation”. So now it’s very hard to find people saying > this is not happening. I think the battle about the recognition of its > existence is kind of being won in many respects. Now people are debating > whether or not it’s a good or a bad idea. > > MG: The other criticism is that this whole fight is being driven by a > right-wing/populist agenda. What would your response be to that? > > Lowenthal: Well, conservatives sense and experience the censorship > probably more than progressives. But that’s only part of the story. A > lot of people and groups on the left are also increasingly censored, for > example around the Israel-Gaza question. So, it impacts anyone who’s > outside of the power core. Right now, it effects the so-called right a > bit more, but it can easily switch in the other direction. And this is > one of the key arguments for progressives who still think this kind of > system of government-sponsored content moderation is a good idea: “One > day they might come for you”. And I think that has certainly already > happened around the Palestine question. > > And as to the “populist” accusation: it is populist, it’s popular voices > that are being censored by elites. That is actually the main thing. It’s > not left or right. It is populist in the sense of “for the people”. It > is a threat to the elites. > > MG: How do you see the situation around censorship and free speech > evolving in Europe and Germany right now, if you follow them from > Australia? > > Lowenthal: Well, Germany from afar it looks quite bad, possibly the > worst in Europe. Germany is extremely important, because of Germany’s > power in the EU and because the cultural work that is leveraged to > justify new systems of censorship (mostly through NGOs and academia) has > a very large presence in Germany. It seems to me that it is the US, then > the UK, then Germany in importance. There have also been victories — > there is a lot more understanding of the censorship-industrial complex > and what it does, and there are more people who are speaking up and out > against it. The Stanford Internet Observatory, a key player in the > complex, is closing down. At the same time the new Labour Government in > Britain plans to repeal the Free Speech Act that was designed to protect > free speech on university campuses and the censors aren’t giving up the > fight easily at all. So, there’s a lot of work still to be done. > > MG: How do you see the situation in Europe and Germany evolving? Where > do you see this going? > > Lowenthal: I think there’s going to be a very big battle. It does not > seem like the elites want to back down from trying to censor the > internet. So, I think, if they’re not willing to leave the internet > open, more and more people are going to mobilise, to demand that the > internet remain free and open. > > [...] The problem, of course, is that many people on the left have > decided that they want the government to get involved in curtailing > speech on the internet. And I think until a significant number of > progressives actually realise just how dangerous that is, it’s going to > be very hard to win this fight and renormalise free speech. The easiest > and practical way to support free speech is to speak freely. The power > is right there in your very own body. > > MG: [...] “Hate is not an opinion”. What’s your response to that? > > Lowenthal: Well, who defines hate? What is hate? And is the government > defining it? Are political parties defining it? What’s the threshold? > What’s the barrier? Wasn’t it Ursula von der Leyen who said: “Hate is > hate”? If that’s the definition, that is a very poor definition. I also > think that repression so often fuels hate speech. If you try and repress > it, it actually grows rather than subsiding. I think there are better > tactics and strategies than censorship. [...] one is just for the > government to be less coercive. Because I think the coercion actually > encourages more hate speech and bad actors. The other one is a very > neutral form of education. The third thing is to pay less attention to > hate speech, because again, like with coercion, attention will actually > foster it. It then becomes a great way to get attention. Like with > children, when they learn that breaking their toys gets them attention, > they will do it more. I think having open conversations is much > better. I understand that a lot of times on the internet that can be > hard. But the other thing is, you can turn the internet off. You don’t > have to read and look at all of this stuff. A lot of it, you can just > ignore. And I think that not creating a moral panic about it is probably > one of the key things, and learning how to have calm and open > conversations across differences. > > --8<---------------cut here---------------end--------------->8--- > > estremamente interessante il parallelo che usa: i bambini che rompono i > giocattoli per ottenere attenzione... > > quindi siamo tutti MINORATI gestiti da innumerevoli tutori > istituzionali, giusto? > > saluti, 380° > > > [1] https://networkaffects.substack.com/ > > [2] https://westminsterdeclaration.org/ >

Aug. 12, 2024

Re: [nexa] “One day they might come for you”

by 380°

Buongiorno, felice di vedere che, seppur faticosamente, pian piano "conspiracy is the new normal", specialmente quando è di stampo POPULISTA :-D perché inutile girarci in giro: Lowenthal è l'archetipo del complottista (a sua insaputa?), con decisi tratti populisti... o no?!? "J.C. DE MARTIN" <juancarlos.demartin(a)polito.it> writes: > *“One day they might come for you” > */Digital rights activist Andrew Lowenthal on progressives' support for > the censorship-industrial complex > / > Thomas Fazi > > Aug 05, 2024 > > In this guest post, Maike Gosch, a German communication strategist, > writer and former lawyer — whose article on the banning of the German > magazine Compact I published recently — interviews Andrew Lowenthal, the > founder and managing director of the digital civil liberties > organisation liber-net. > > Lowenthal is an Australian digital rights activist of German-Jewish > descent. For almost 18 years he was the Executive Director of > EngageMedia, an Asia-based NGO focused on human rights online, freedom > of expression and open technology. The digital rights environment in > which Lowenthal spent most of his adult life was avowedly progressive — > as was Lowenthal himself. > continua qui: https://www.thomasfazi.com/p/one-day-they-might-come-for-you siccome non vorrei che i lettori si perdessero il nocciolo della questione, ovvero la metamorfosi del progressismo di sinistra, l'articolo continua così: --8<---------------cut here---------------start------------->8--- Then came the pandemic, and Lowenthal, as many of us, experienced an anthological shock that led him to reconsider much of his worldview — especially about his activist milieu and the progressive left more in general. [...] Increasingly estranged from the progressive former-digital-rights-now-turned-pro-censorship movement, Lowenthal began to focus his work on the emerging digital censorship regime — or censorship-industrial complex — even collaborating with journalists like Matt Taibi and Michael Shellenberger on the Twitter Files. This eventually led to the founding of liber-net. To stay on top of all things online censorship-related, make sure to follow Lowenthal’s great Substack, Network Affects [1]. Lowenthal: [...] got very interested in left-wing activism [...] and particularly what was known was “anti-globalisation”. [...] I helped organise, in Melbourne against the World Economic Forum [...] then I started liber-net, which focuses on digital civil liberties and digital authoritarianism in the West, issues including censorship, free speech, surveillance, ways in which technologies increasingly create digital gates and barriers around access, such as what we saw during Covid with the vaccine passports, that are now coming more, in terms of digital ID issues around programmable currency, so essentially systems of digital social control. Lowenthal: [...] (sulla Westminster Declaration [2], n.d.r.) the media on the other hand rather dismissed it, or they just focus on the right-wing people who signed it and ignored the left-wing people who signed it to try and suggest that it was very bad, and you should stay away from it. MG: Often the reaction in Germany when people hear about the censorship-industrial complex is that it gets called a conspiracy theory. What would your reply be to that? What do you think of that criticism? Lowenthal: There’s huge amount of evidence now that there are a massive number of organisations coordinating to censor content on the internet. In fact, a lot of liberal and progressive groups now acknowledge that this is happening. At the start of the Twitter Files, the line was, “Oh, no, it’s a conspiracy theory. None of this is happening, etc.”. Now, most people are saying: “Well, yes, the government and NGOs are involved in content moderation. But it’s okay. It’s legal. Governments should be paying attention and protecting people from misinformation”. So now it’s very hard to find people saying this is not happening. I think the battle about the recognition of its existence is kind of being won in many respects. Now people are debating whether or not it’s a good or a bad idea. MG: The other criticism is that this whole fight is being driven by a right-wing/populist agenda. What would your response be to that? Lowenthal: Well, conservatives sense and experience the censorship probably more than progressives. But that’s only part of the story. A lot of people and groups on the left are also increasingly censored, for example around the Israel-Gaza question. So, it impacts anyone who’s outside of the power core. Right now, it effects the so-called right a bit more, but it can easily switch in the other direction. And this is one of the key arguments for progressives who still think this kind of system of government-sponsored content moderation is a good idea: “One day they might come for you”. And I think that has certainly already happened around the Palestine question. And as to the “populist” accusation: it is populist, it’s popular voices that are being censored by elites. That is actually the main thing. It’s not left or right. It is populist in the sense of “for the people”. It is a threat to the elites. MG: How do you see the situation around censorship and free speech evolving in Europe and Germany right now, if you follow them from Australia? Lowenthal: Well, Germany from afar it looks quite bad, possibly the worst in Europe. Germany is extremely important, because of Germany’s power in the EU and because the cultural work that is leveraged to justify new systems of censorship (mostly through NGOs and academia) has a very large presence in Germany. It seems to me that it is the US, then the UK, then Germany in importance. There have also been victories — there is a lot more understanding of the censorship-industrial complex and what it does, and there are more people who are speaking up and out against it. The Stanford Internet Observatory, a key player in the complex, is closing down. At the same time the new Labour Government in Britain plans to repeal the Free Speech Act that was designed to protect free speech on university campuses and the censors aren’t giving up the fight easily at all. So, there’s a lot of work still to be done. MG: How do you see the situation in Europe and Germany evolving? Where do you see this going? Lowenthal: I think there’s going to be a very big battle. It does not seem like the elites want to back down from trying to censor the internet. So, I think, if they’re not willing to leave the internet open, more and more people are going to mobilise, to demand that the internet remain free and open. [...] The problem, of course, is that many people on the left have decided that they want the government to get involved in curtailing speech on the internet. And I think until a significant number of progressives actually realise just how dangerous that is, it’s going to be very hard to win this fight and renormalise free speech. The easiest and practical way to support free speech is to speak freely. The power is right there in your very own body. MG: [...] “Hate is not an opinion”. What’s your response to that? Lowenthal: Well, who defines hate? What is hate? And is the government defining it? Are political parties defining it? What’s the threshold? What’s the barrier? Wasn’t it Ursula von der Leyen who said: “Hate is hate”? If that’s the definition, that is a very poor definition. I also think that repression so often fuels hate speech. If you try and repress it, it actually grows rather than subsiding. I think there are better tactics and strategies than censorship. [...] one is just for the government to be less coercive. Because I think the coercion actually encourages more hate speech and bad actors. The other one is a very neutral form of education. The third thing is to pay less attention to hate speech, because again, like with coercion, attention will actually foster it. It then becomes a great way to get attention. Like with children, when they learn that breaking their toys gets them attention, they will do it more. I think having open conversations is much better. I understand that a lot of times on the internet that can be hard. But the other thing is, you can turn the internet off. You don’t have to read and look at all of this stuff. A lot of it, you can just ignore. And I think that not creating a moral panic about it is probably one of the key things, and learning how to have calm and open conversations across differences. --8<---------------cut here---------------end--------------->8--- estremamente interessante il parallelo che usa: i bambini che rompono i giocattoli per ottenere attenzione... quindi siamo tutti MINORATI gestiti da innumerevoli tutori istituzionali, giusto? saluti, 380° [1] https://networkaffects.substack.com/ [2] https://westminsterdeclaration.org/ -- 380° (Giovanni Biscuolo public alter ego) «Noi, incompetenti come siamo, non abbiamo alcun titolo per suggerire alcunché» Disinformation flourishes because many people care deeply about injustice but very few check the facts. Ask me about <https://stallmansupport.org>.

Aug. 12, 2024

Re: [nexa] Large Language Models pose a risk to society and need tighter regulation, say Oxford researchers

by Marco Fioretti

On Sun, Aug 11, 2024 at 14:24, Alberto Cammozzo via nexa wrote: Come ha già detto Quintarelli parlando di "scala, velocità e rimedio", il punto credo sia non la veridicità ma le conseguenze, la scala, gli scopi della diffusione, e soprattutto l'assunzione di responsabilità di chi diffonde affermazioni. Se un influencer con zilioni di follower dice che il suo unicorno ha la gotta, o che la terra è piatta, nessuno lo porterà in tribunale. Se lo stesso diffonde notizie false che scatenano un pogrom, sì, e giustamente, specie se si può dimostrare che lo scopo era proprio quello. Se la stessa notizia falsa e pericolosa la diffonde uno che ha tre follower, nessuno dei quali fa nulla, nessuno si prenderà la briga di perseguirlo. per la cronaca, questo e' il motivo per cui anni fa proponevo, come soluzione parziale ma necessaria, di limitare "the amount of time during which they can publish every day" https://stop.zona-m.net/2021/01/we-really-need-two-minutes-of-hate-every-da… (https://stop.zona-m.net/2021/01/we-really-need-two-minutes-of-hate-every-da…) https://stop.zona-m.net/2021/10/censorship-or...-efficient-slowness/ (https://stop.zona-m.net/2021/10/censorship-or...-efficient-slowness/) Marco

Aug. 12, 2024

Re: [nexa] Large Language Models pose a risk to society and need tighter regulation, say Oxford researchers

by Giacomo Tesio

Caro Guido, l'importante, in uno stato di diritto è che la Legge sia uguale per tutti e nessuno possa sperare di sottrarvisi attraverso automatismi che compiono reati per proprio conto. Il 11 Agosto 2024 07:12:04 UTC, Guido Vetere <vetere.guido(a)gmail.com> ha scritto: > il crimine non consiste nel dare > credito a una affermazione non verificata (altrimenti staremmo tutti in > galera), e neanche raccontare balle al telefono. il crimine consiste nel > diffondere attraverso mezzi di comunicazione di massa notizie non > verificate che possono fomentare la violenza. Se questo fosse il crimine, bisognerebbe arrestare la maggioranza dei giornalisti dall'inizio della guerra in Ucraina in poi. Non capisco poi la fretta di assolvere i mezzi di comunicazione di massa (giornali, TV, LLM distribuiti gratuitamente...) e quelli di manipolazione individuale su larga scala (social vari, LLM forniti come SaaS etc...): ad esempio, nel caso concreto, è stato Musk a decidere di inviare la fake news a persone con probabilità elevata avrebbero reagito violentemente. Il fatto che l'abbia fatto tramite un software (Twitter) deve essere irrilevante. > fuor di metafora: l'idea che i LLM debbano essere per legge tenuti alla > veridicità è... una cazzata. Siamo d'accordo: i LLM sono espressione di chi li realizza, distribuisce e amministra. Non si può pretendere di imporre una verità a queste persone più di quanto non la si possa imporre a qualsiasi altro cittadino che si esprima in altro modo, attraverso l'arte, la poesia, la letteratura, il giornalismo etc... Semplicemente bisogna sempre considerarli direttamente responsabili del loro output, esattamente come se lo avessero scritto di proprio pugno su un giornale a tiratura planetaria. Niente di più e niente di meno. Giacomo

Aug. 11, 2024

Re: [nexa] Large Language Models pose a risk to society and need tighter regulation, say Oxford researchers

by Alberto Cammozzo

Come ha già detto Quintarelli parlando di "scala, velocità e rimedio", il punto credo sia non la veridicità ma le conseguenze, la scala, gli scopi della diffusione, e soprattutto l'assunzione di responsabilità di chi diffonde affermazioni. Se un influencer con zilioni di follower dice che il suo unicorno ha la gotta, o che la terra è piatta, nessuno lo porterà in tribunale. Se lo stesso diffonde notizie false che scatenano un pogrom, sì, e giustamente, specie se si può dimostrare che lo scopo era proprio quello. Se la stessa notizia falsa e pericolosa la diffonde uno che ha tre follower, nessuno dei quali fa nulla, nessuno si prenderà la briga di perseguirlo. Se il mega influencer diffonde notizie -anche vere- che hanno conseguenze per terzi dovrà essere in grado di dimostrarne la fondatezza, forse anche in tribunale se questi si ritengono lesi o diffamati; quindi: conseguenze. Che abbia preso le notizie dal blog di suo cugino, da un motore di ricerca, da un oroscopo o da un LLM poco conta. Questo per quanto riguarda le persone: ci sia aspetta che un umano sia in grado di rispondere delle conseguenze del proprio output, e della veridicità solo in conseguenza. I software al contrario vantano uno statuto di sostanziale irresponsabilità sancito consuetudianariamente dai /terms of use/, per cui la responsabilità dell'uso del loro output cade su chi li usa e non su chi produce il software. Il problema con gli LLM (condiviso con i social) è che sono emanazione di società che usano software per produrre/diffondere affermazioni: (1) per lucro, (2) massimizzando la scala di diffusione industrialmente e (3) minimizzando la propria responsabilità, estendendo a proprio vantaggio l'irresponsabilità del software. Come per i social, non è irragionevole che un business che lucra sulla diffusione di affermazioni si assuma la responsabilità sulle conseguenze della diffusione delle stesse, considerando la diffusione di notizie false non tanto per la loro falsità, quanto per la scala e le potenziali conseguenze della loro diffusione. Ed eventualmente per il sussistere di scopi criminosi, ai quali incita terzi. Ma per favore non diciamo che il tal LLM o il tal social (o internet o il web) dice questo o quello. Quello che viene detto viene detto o da persone o da imprese, che antropomorfizzano un software irresponsabile per non assumersi la responsabilità delle conseguenze del potere che hanno grazie a tal software. E che vendono. Buona domenica. Alberto On 11/08/24 09:12, Guido Vetere wrote: > altro esperimento mentale, della serie: "se voi foste il giudice" > (valida alternativa alle parole crociate sulla settimana enigmistica) > > la signora di Chester, alla sbarra per istigazione all'odio sul web, > dice al giudice che la bufala sul musulmano clandestino gliel'aveva > raccontata suo cugino al telefono. > > quesiti: si potrebbe trattare di una attenuante? in tal caso, sarebbe > ancora un'attenuante se la signora fosse stata al corrente che il > cugino è un paranoico? infine: bisognerebbe incriminare anche il cugino? > > se io fossi il giudice, direi di no: il crimine non consiste nel dare > credito a una affermazione non verificata (altrimenti staremmo tutti > in galera), e neanche raccontare balle al telefono. il crimine > consiste nel diffondere attraverso mezzi di comunicazione di massa > notizie non verificate che possono fomentare la violenza. la paranoia > conclamata del cugino non rileva. (i legali mi correggano) > > fuor di metafora: l'idea che i LLM debbano essere per legge tenuti > alla veridicità è da una parte ingenua (veridicità secondo quale > episteme? ah saperlo!), dall'altra inutile: la decisione cruciale > riguarda come usare l'AI generativa, ed è una decisione tutta umana. > pensare il contrario significa abbracciare una visione pericolosamente > tecno-positivista del problema etico della conoscenza, ben noto fin > dai tempi antichi. > > buona domenica! > G. > > > > > > > > > > On Sat, 10 Aug 2024 at 17:59, Giacomo Tesio <giacomo(a)tesio.it> wrote: > > Salve Guido, > > Il 10 Agosto 2024 08:36:32 UTC, Guido Vetere > <vetere.guido(a)gmail.com> ha scritto: > > Ora un esperimento mentale: si scopre che la mentitrice ha > generato il > > falso per mezzo di un LLM. > > Quesito: potrà usare questa circostanza a sua, sia pur parziale, > discolpa? > > Dipende. > > Era consapevole di come funziona un LLM? > Era consapevole della falsità dell'interpretazione che gli altri > avrebbero attribuito alla > sequenza di caratteri calcolati dal > software? > > Se no, non si potrebbe attribuirgli alcuna responsabilità, che > sarebbe SOLO di coloro > che le hanno fornito accesso al LLM in questione. > > Se sì, la responsabilità delle conseguenze della diffusione di > tale output ricadrebbe > anche su di lei. > > > Se sì, bisognerà dare ragione ai ricercatori di Oxford, > altrimenti avranno > > ricercato invano. > > Non sequitur: i ricercatori di Oxford possono aver "ricercato > invano" anche se la > la tua ipotetica signora fosse vittima della malafede di chi > realizza e promuove > come "intelligenze artificiali" questi software programmati > statisticamente. > > Non si tratta di istituire "verità" da imporre ai software (o > meglio, tramite i software). > Si tratta semplicemente di individuare i responsabili, senza > permettere > a nessuno di nascondersi dietro un software, che si chiami Twitter > o ChatGPT. > > Che si tratti della 56 enne arrestata a Chester, di Musk. di > Altman, di Bankman-Fried... > > > Giacomo >

Aug. 11, 2024

Re: [nexa] Large Language Models pose a risk to society and need tighter regulation, say Oxford researchers

by Guido Vetere

altro esperimento mentale, della serie: "se voi foste il giudice" (valida alternativa alle parole crociate sulla settimana enigmistica) la signora di Chester, alla sbarra per istigazione all'odio sul web, dice al giudice che la bufala sul musulmano clandestino gliel'aveva raccontata suo cugino al telefono. quesiti: si potrebbe trattare di una attenuante? in tal caso, sarebbe ancora un'attenuante se la signora fosse stata al corrente che il cugino è un paranoico? infine: bisognerebbe incriminare anche il cugino? se io fossi il giudice, direi di no: il crimine non consiste nel dare credito a una affermazione non verificata (altrimenti staremmo tutti in galera), e neanche raccontare balle al telefono. il crimine consiste nel diffondere attraverso mezzi di comunicazione di massa notizie non verificate che possono fomentare la violenza. la paranoia conclamata del cugino non rileva. (i legali mi correggano) fuor di metafora: l'idea che i LLM debbano essere per legge tenuti alla veridicità è da una parte ingenua (veridicità secondo quale episteme? ah saperlo!), dall'altra inutile: la decisione cruciale riguarda come usare l'AI generativa, ed è una decisione tutta umana. pensare il contrario significa abbracciare una visione pericolosamente tecno-positivista del problema etico della conoscenza, ben noto fin dai tempi antichi. buona domenica! G. On Sat, 10 Aug 2024 at 17:59, Giacomo Tesio <giacomo(a)tesio.it> wrote: > Salve Guido, > > Il 10 Agosto 2024 08:36:32 UTC, Guido Vetere <vetere.guido(a)gmail.com> ha > scritto: > > Ora un esperimento mentale: si scopre che la mentitrice ha generato il > > falso per mezzo di un LLM. > > Quesito: potrà usare questa circostanza a sua, sia pur parziale, > discolpa? > > Dipende. > > Era consapevole di come funziona un LLM? > Era consapevole della falsità dell'interpretazione che gli altri avrebbero > attribuito alla > sequenza di caratteri calcolati dal > software? > > Se no, non si potrebbe attribuirgli alcuna responsabilità, che sarebbe > SOLO di coloro > che le hanno fornito accesso al LLM in questione. > > Se sì, la responsabilità delle conseguenze della diffusione di tale output > ricadrebbe > anche su di lei. > > > Se sì, bisognerà dare ragione ai ricercatori di Oxford, altrimenti > avranno > > ricercato invano. > > Non sequitur: i ricercatori di Oxford possono aver "ricercato invano" > anche se la > la tua ipotetica signora fosse vittima della malafede di chi realizza e > promuove > come "intelligenze artificiali" questi software programmati > statisticamente. > > Non si tratta di istituire "verità" da imporre ai software (o meglio, > tramite i software). > Si tratta semplicemente di individuare i responsabili, senza permettere > a nessuno di nascondersi dietro un software, che si chiami Twitter o > ChatGPT. > > Che si tratti della 56 enne arrestata a Chester, di Musk. di Altman, di > Bankman-Fried... > > > Giacomo >

Aug. 11, 2024

AI as a tool to locate document for Foia request

by Benedetto Ponti

Reasonable for Cabinet Office not to use AI, even if it might save time Tribunal finds that it was reasonable for the Cabinet Office not to use AI to locate documents for FOIA Request Overview of the FTT's decision on AI use in FOIA requests In a recent decision of the First-tier Tribunal (General Regulatory Chamber) Information Rights (FTT), the Cabinet Office succeeded in its application for a strike out. An individual, Mr Ed Surridge (the Appellant) had appealed against a Decision Notice (DN) issued by the Information Commissioner. The Cabinet Office said that it would take them too long to locate the information he had requested. Mr Surridge appealed, and argued that the Cabinet Office should use artificial intelligence (AI) tools to locate the requested information. The Tribunal held that the Cabinet Office's position was reasonable given it did not have the AI tools to perform the search. FOIA request and appeal process The appeal concerned a request for information under the Freedom of Information Act 2000 (FOIA) about wildfire preparedness for the safeguarding of people and their homes near woodland. The Cabinet Office refused the request on the basis that it would exceed the appropriate limit of £600 or 24 hours of work under section 12(1) FOIA and the Fees Regulations. The scope of the request was for a review of over 3,568 emails which the Cabinet Office estimated would take 43 hours to accomplish. The Appellant complained to the Commissioner, who issued a DN which upheld the Cabinet Office's decision. The Appellant then appealed the DN in the FTT, challenging the adequacy of the searches conducted by the Cabinet Office, and suggesting that AI tools could be used to locate the information more efficiently. The Cabinet Office applied to strike out the appeal. It argued that there was no prospect of the FTT finding an error of law in the Information Commissioner's judgment. The Applicant's views on AI search methods had no prospect of showing an error of law in the section 12 decision; AI chatbot responses could not be relied upon; and the Cabinet Office did not have AI tools to assist in the searches. Tribunal decision and rejection of AI search based argument The Tribunal granted the strike out application on the ground that the appeal had "no reasonable prospect of success" under 8(3)(c) of the Tribunal Rules. It assessed the applicability of the section 12 FOIA limit based on the processes "actually used and adopted" by the public authority, not on how the information "should have been kept" or searched. The Tribunal accepted the Cabinet Office's evidence that it did not have or use AI tools so therefore could not search its records "in the way that the Appellant envisaged." The Tribunal also referred to Oakley v Information Commissioner [2024] UKFTT 00315 (GRC) to highlight the inadequacy of AI searches. AI evidence was compared to expert evidence noting that "an expert would be required to explain their expertise, the sources that they rely upon and the methodology that they applied." Therefore "little weight" was given to the Appellant's evidence based on AI searches, as no explanation of the sources or methodology used by the AI tool could be provided. How might this ruling shape the future of the use of AI in public sector information management? The FTT's decision highlights the limitations of using AI within government processes in the UK. Public authorities are not obliged to utilise AI tools if they are not already applied in their operations and it would not be appropriate to mandate this - even if such tools may theoretically enhance the efficiency of workflow. As AI technologies continue to evolve, public authorities face the decision of reassessing their information management practices and consider integrating AI tools to enhance their ability to respond to information requests efficiently. A decision to integrate AI tools into the daily processes of government activities would come with its own considerations and challenges, such as dealing with the lack of transparency in the methodology of AI-generated evidence. Addressing such concerns is crucial for effective implementation in the long term. This decision shows that there is no positive obligation to do so, yet. https://www.lexology.com/library/detail.aspx?g=cc3fb15f-34f3-4517-9c78-3292… Inviato da Outlook per Android<https://aka.ms/AAb9ysg>

Aug. 10, 2024

Re: [nexa] Large Language Models pose a risk to society and need tighter regulation, say Oxford researchers

by Giacomo Tesio

Salve Guido, Il 10 Agosto 2024 08:36:32 UTC, Guido Vetere <vetere.guido(a)gmail.com> ha scritto: > Ora un esperimento mentale: si scopre che la mentitrice ha generato il > falso per mezzo di un LLM. > Quesito: potrà usare questa circostanza a sua, sia pur parziale, discolpa? Dipende. Era consapevole di come funziona un LLM? Era consapevole della falsità dell'interpretazione che gli altri avrebbero attribuito alla sequenza di caratteri calcolati dal software? Se no, non si potrebbe attribuirgli alcuna responsabilità, che sarebbe SOLO di coloro che le hanno fornito accesso al LLM in questione. Se sì, la responsabilità delle conseguenze della diffusione di tale output ricadrebbe anche su di lei. > Se sì, bisognerà dare ragione ai ricercatori di Oxford, altrimenti avranno > ricercato invano. Non sequitur: i ricercatori di Oxford possono aver "ricercato invano" anche se la la tua ipotetica signora fosse vittima della malafede di chi realizza e promuove come "intelligenze artificiali" questi software programmati statisticamente. Non si tratta di istituire "verità" da imporre ai software (o meglio, tramite i software). Si tratta semplicemente di individuare i responsabili, senza permettere a nessuno di nascondersi dietro un software, che si chiami Twitter o ChatGPT. Che si tratti della 56 enne arrestata a Chester, di Musk. di Altman, di Bankman-Fried... Giacomo

Aug. 10, 2024

“One day they might come for you”

by J.C. DE MARTIN

*“One day they might come for you” */Digital rights activist Andrew Lowenthal on progressives' support for the censorship-industrial complex / Thomas Fazi Aug 05, 2024 In this guest post, Maike Gosch, a German communication strategist, writer and former lawyer — whose article on the banning of the German magazine Compact I published recently — interviews Andrew Lowenthal, the founder and managing director of the digital civil liberties organisation liber-net. Lowenthal is an Australian digital rights activist of German-Jewish descent. For almost 18 years he was the Executive Director of EngageMedia, an Asia-based NGO focused on human rights online, freedom of expression and open technology. The digital rights environment in which Lowenthal spent most of his adult life was avowedly progressive — as was Lowenthal himself. [...] continua qui: https://www.thomasfazi.com/p/one-day-they-might-come-for-you

Aug. 10, 2024