July 2025 - nexa - server-nexa.polito.it

[Responsible AI Talk] The AI Alignment Problem, Camilla Bignotti, Banca d’Italia

by Daniele Quercia

The AI Alignment Problem: Bridging Lab Research and Policy Solutions Camilla Bignotti, Banca d’Italia Join the meeting <https://teams.microsoft.com/l/meetup-join/19%3ameeting_MDc4MGUzZWMtY2FhNi00…> The rapid advancement of large language models (LLMs) raises significant concerns about their alignment with human values and goals, particularly given their potential impact on individual lives. The alignment problem can be analysed through two complementary perspectives: conducting empirical experimentation to assess whether AI models align with human expectations or designing policy instruments to implement effective AI governance. This seminar will combine both approaches. Starting from an empirical analysis of how large language models interpret constitutional principles in complex decision-making scenarios, we demonstrate the risks posed by polarized or biased models in complex decision-making contexts. These findings highlight the need for robust training and evaluation methods to address these challenges and ensure model alignment with societal values. This challenge extends beyond the specific domain explored in our study—it represents a crucial aspect related to foundational models applications across their diverse domains. Therefore, we must investigate comprehensive measures to tackle the alignment problem while promoting safe and responsible AI development without stifling innovation. Public-private partnerships are fundamental building blocks for effective AI governance. By creating spaces for dialogue on standard-setting and pilot testing, regulatory institutions and private actors can align their objectives toward responsible innovation. Innovation facilitators established by several European Central Banks exemplify effective tools for promoting stakeholder cooperation in responsible AI deployment. Milano Hub, Banca d'Italia's innovation centre, provides a compelling use case: through its Call for Proposals, it supports promising solutions while steering innovation in a positive direction. <https://arxiv.org/abs/2407.02191> Camilla Bignotti serves as a legal expert at Banca d'Italia's Innovation Centre "Milano Hub," where she contributes to central banking activities that support the digital transformation of the banking, financial, and payments sectors. Her work includes fostering international cooperation among regulatory institutions and innovation hubs in the field of technological evolution. Recently, she was seconded to the Central Bank of Ireland to collaborate on designing a new regulatory sandbox program. Camilla is a qualified lawyer with a background in transnational and comparative law and growing expertise in regulatory frameworks related to financial innovation and artificial intelligence. Subscribe to future talk announcements: Anyone outside Bell Labs can receive talk announcements by subscribing to the mailing list. To subscribe, send an empty email with the subject line "Subscribe RAI" to daniele.quercia(a)nokia-bell-labs.com<mailto:daniele.quercia@nokia-bell-labs.com>.

July 28, 2025

Re: [nexa] "China releases AI action plan days after the U.S. as global tech race heats up"

by Giuseppe Attardi

July 28, 2025

"China releases AI action plan days after the U.S. as global tech race heats up"

by J.C. DE MARTIN

China releases AI action plan days after the U.S. as global tech race heats up Published Sat, Jul 26 2025 Evelyn Cheng <https://www.cnbc.com/evelyn-cheng/>@in/evelyn-cheng-53b23624 <https://linkedin.com/in/evelyn-cheng-53b23624>@chengevelyn <https://twitter.com/chengevelyn> Key Points * China on Saturday released a global action plan for artificial intelligence that called for international cooperation on tech development and regulation. * Premier Li Qiang announced China has proposed the establishment of a global AI cooperation organization, according to an official readout. * Former Google CEO Eric Schmidt met with Shanghai Party Secretary Chen Jining on Thursday in the city ahead of the AI conference, according to a city announcement. Continua qui: https://www.cnbc.com/2025/07/26/china-ai-action-plan.html

July 27, 2025

Re: [nexa] The first copyright ruling on generative AI training is a win for AI labs

by Maurizio Borghi

Sì è corretto. Il giudice distingue tra la modalità di acquisizione delle opere protette e il loro effettivo impiego nel training dell'IA. Se le opere sono ottenute legalmente, il loro utilizzo a fini di addestramento costituisce "fair use"; se invece sono ottenute illegalmente (ovvero da database "pirata"), l'utilizzo non è ammesso. Teniamo però presente che si tratta della prima sentenza in materia e altre pronunce seguiranno. Non è detto che la giurisprudenza USA confermerà questo orientamento. Un caro saluto, MB Il giorno ven 25 lug 2025 alle ore 19:54 Enrico Nardelli < nardelli(a)mat.uniroma2.it> ha scritto: > (non mi pare la notizia sia passata in lista) > > https://www.understandingai.org/p/the-first-copyright-ruling-on-generative > > In realtà, se capisco bene, una vittoria parziale, perché il giudice ha > stabilito che l'uso di copie pirata dei libri per addestrare il suo LLM > fatto da Anthropic è illegale, ma solo perché le copie dei libri sono > pirata. > > Quindi, sempre se capisco bene, qualora un'azienda invece i libri usati > per addestrare un LLM li compri, allora l'addestramento non ha necessità di > richiedere il consenso di chi detiene i diritti dei libri stessi, perché è > un utilizzo trasformativo. > > Sbaglio/trascuro qualcosa? > > Grazie in anticipo, Enrico > -- > > -- EN > https://www.hoepli.it/libro/la-rivoluzione-informatica/9788896069516.html > ====================================================== > Prof. Enrico Nardelli > Past President di "Informatics Europe" > Direttore del Laboratorio Nazionale "Informatica e Scuola" del CINI > Dipartimento di Matematica - Università di Roma "Tor Vergata" > Via della Ricerca Scientifica snc - 00133 Roma > home page: https://www.mat.uniroma2.it/~nardelli > blog: https://link-and-think.blogspot.it/ > tel: +39 06 7259.4204 fax: +39 06 7259.4699 > mobile: +39 335 590.2331 e-mail: nardelli(a)mat.uniroma2.it > online meeting: https://blue.meet.garr.it/b/enr-y7f-t0q-ont > ====================================================== > -- > -- _______________ *Maurizio Borghi* Università di Torino https://www.dg.unito.it/persone/maurizio.borghi Co-Director Nexa Center for Internet & Society <https://nexa.polito.it/>

July 26, 2025

Zitron, "The Hater's Guide To The AI Bubble"

by J.C. DE MARTIN

*The Hater's Guide To The AI Bubble* Edward Zitron Jul 21, 2025 55 min read Good journalism is making sure that history is actively captured and appropriately described and assessed, and it's accurate to describe things as they currently are as alarming. And I am /alarmed./ Alarm is not a state of weakness, or belligerence, or myopia. My concern does not dull my vision, even though it's convenient to frame it as somehow alarmist, like I have some hidden agenda or bias toward doom. I profoundly dislike the financial waste, the environmental destruction, and, fundamentally, I dislike the attempt to gaslight people into swearing fealty to a sickly and frail psuedo-industry where everybody but NVIDIA and consultancies lose money. I also dislike the fact that I, and others like me, are held to a remarkably different standard to those who paint themselves as "optimists," which typically means "people that agree with what the market wishes were true." Critics are continually badgered, prodded, poked, mocked, and jeered at for not automatically aligning with the idea that generative AI will be this massive industry, constantly having to prove themselves, as if somehow there's something malevolent or craven about criticism, that critics "do this for clicks" or "to be a contrarian." I don't do anything for clicks. I don't have any stocks or short positions. My agenda is simple: I like writing, it comes to me naturally, I have a podcast, and it is, on some level, my job to try and understand what the tech industry is doing on a day-to-day basis. It is easy to try and dismiss what I say as going against the grain because "AI is big," but I've been railing against bullshit bubbles since 2021 — the_anti-remote work push_ <https://www.wheresyoured.at/the-upcoming-remote-work-company/> (_and the people behind it_ <https://www.wheresyoured.at/the-big-media-anti-remote-op-ed-complex/>),_the Clubhouse and audio social networks bubble_ <https://www.wheresyoured.at/clubhouse-is-the-big-stinker-that/>,_the NFT bubble_ <https://www.wheresyoured.at/the-nihilism-and-exploitation-of/>,_the made-up quiet quitting panic_ <https://www.wheresyoured.at/quiet-quitting-and-the-death-of-office/>, and I even, though not as clearly as I wished,_called that something was up with FTX several months before it imploded_ <https://www.wheresyoured.at/sam-bankrun-fraud/>. This isn't "contrarianism." It's the kind of skepticism of power and capital that's necessary to meet these moments, and if it's necessary to dismiss my work because it makes you feel icky inside, get a therapist or see a priest. Nevertheless, I am /alarmed/, and while I have said some of these things separately, based on recent developments, I think it's necessary to say why. In short, I believe the AI bubble is deeply unstable, built on vibes and blind faith, and when I say "the AI bubble," I mean the entirety of the AI trade. And it's alarmingly simple, too. [...] continua qui: https://www.wheresyoured.at/the-haters-gui/

July 26, 2025

La Danimarca inizia la battaglia contro i deepfake

by Diego Giorio

Buongiorno a tutti. Non mi sembra che la notizia sia già passata in lista e credo possa essere di interesse https://privacyitaliana.com/2025/07/25/inizia-la-battaglia-contro-i-deepfak… Buon fine settimana a tutti D.

July 26, 2025

Re: [nexa] The first copyright ruling on generative AI training is a win for AI labs

by Stefano Quintarelli

quella è una prima parte, da quanto capisco puoi allenare un modello con testi che compri, di cui fai scansioni ed ocr e mi sembra ragionevole nulla è detto se però, poi, invece di usarlo per fini personali eroghi un servizio a terzi... Il 25 luglio 2025 17:54:36 UTC, Enrico Nardelli <nardelli(a)mat.uniroma2.it> ha scritto: >(non mi pare la notizia sia passata in lista) > >https://www.understandingai.org/p/the-first-copyright-ruling-on-generative > >In realtà, se capisco bene, una vittoria parziale, perché il giudice ha stabilito che l'uso di copie pirata dei libri per addestrare il suo LLM fatto da Anthropic è illegale, ma solo perché le copie dei libri sono pirata. > >Quindi, sempre se capisco bene, qualora un'azienda invece i libri usati per addestrare un LLM li compri, allora l'addestramento non ha necessità di richiedere il consenso di chi detiene i diritti dei libri stessi, perché è un utilizzo trasformativo. > >Sbaglio/trascuro qualcosa? > >Grazie in anticipo, Enrico >

July 25, 2025

The first copyright ruling on generative AI training is a win for AI labs

by Enrico Nardelli

(non mi pare la notizia sia passata in lista) https://www.understandingai.org/p/the-first-copyright-ruling-on-generative In realtà, se capisco bene, una vittoria parziale, perché il giudice ha stabilito che l'uso di copie pirata dei libri per addestrare il suo LLM fatto da Anthropic è illegale, ma solo perché le copie dei libri sono pirata. Quindi, sempre se capisco bene, qualora un'azienda invece i libri usati per addestrare un LLM li compri, allora l'addestramento non ha necessità di richiedere il consenso di chi detiene i diritti dei libri stessi, perché è un utilizzo trasformativo. Sbaglio/trascuro qualcosa? Grazie in anticipo, Enrico -- -- EN https://www.hoepli.it/libro/la-rivoluzione-informatica/9788896069516.html ====================================================== Prof. Enrico Nardelli Past President di "Informatics Europe" Direttore del Laboratorio Nazionale "Informatica e Scuola" del CINI Dipartimento di Matematica - Università di Roma "Tor Vergata" Via della Ricerca Scientifica snc - 00133 Roma home page: https://www.mat.uniroma2.it/~nardelli blog: https://link-and-think.blogspot.it/ tel: +39 06 7259.4204 fax: +39 06 7259.4699 mobile: +39 335 590.2331 e-mail: nardelli(a)mat.uniroma2.it online meeting: https://blue.meet.garr.it/b/enr-y7f-t0q-ont ====================================================== --

July 25, 2025

Explanatory Notice and Template for the Public Summary of Training Content for general-purpose AI models

by Antonio Vetro'

Pubblicato oggi dalla Commissione Europea lo schema per la documentazione dei dati di training modelli c.d. “general-purpose" https://digital-strategy.ec.europa.eu/en/library/explanatory-notice-and-tem…

July 24, 2025

Rapporto AAUP, "Artificial Intelligence and Academic Professions"

by J.C. DE MARTIN

*Artificial Intelligence and Academic Professions * American Association of University Professors https://www.aaup.org/reports-publications/aaup-policies-reports/topical-rep… Executive Summary Educational technology, or /ed-tech/, including artificial intelligence (AI), continues to become more integrated into teaching and research in higher education, with minimal oversight. The AAUP’s ad hoc Committee on Artificial Intelligence and Academic Professions—composed of higher education faculty members, staff, and scholars interested in technology and its impact on academic labor—was formed under the assumption that faculty members are best positioned to understand and improve teaching and learning conditions, including the development and implementation of institutional policies around educational technology. To learn more about the experiences and priorities of AAUP members, the committee conducted a survey with a sample of five hundred members from nearly two hundred campuses across the country, collected during a two-week time period. Respondents emphasized the importance of improving education on AI, promoting shared governance through policies and oversight, and focusing on equity, transparency, and worker protections. Based on those responses, the committee identified the five key concerns listed below and described more fully in the findings section of this report. 1. Improving Professional Development Regarding AI and Technology Harms * Despite the widespread use of ed-tech, there is an overall lack of understanding about the relationship between AI and commonly used data-intensive educational technologies. * Untested and unproven technologies are adopted uncritically 2. Implementing Shared Governance Policies to Promote Oversight * AI integration initiatives are spearheaded by administrations with little input from faculty members and other campus community members, including staff and students. * High levels of concern arose around AI and technology procurement, deployment, and use; dehumanized relations; and poor working and learning conditions. 3. Improving Working and Learning Conditions * Preexisting work intensification and devaluation are the main reasons respondents give for using AI to assist with academic tasks. * Implementing AI in higher education adds to faculty and staff workloads and exacerbates long-standing inequities. * AI raises concerns around bias, discrimination, and accessibility because of the untested and uneven impacts on students and student learning. 4. Demanding Transparency and the Ability to Opt Out * Faculty members and staff lack choice and meaningful avenues to opt out of both AI-based tools and other ed-tech. * Few institutions have created transparent, equitable policies or provided effective professional development opportunities on AI use. 5. Protecting Faculty Members and Other Academic Workers * Academic workers across job categories are worried about increased reliance on contingent appointments and declining wages. * Respondents expressed concern about academic freedom and intellectual property rights. The report provides details on the survey’s findings about these concerns and addresses them with recommendations to improve higher education—both broadly and narrowly as it relates to emerging technologies. Faculty members can work to implement these recommendations on their campuses by incorporating guidelines in faculty handbooks and collective bargaining agreements. The recommendations can inform strategy for organizing and policymaking related to AI in higher education institutions and organized labor more generally. The ad hoc Committee on Artificial Intelligence and Academic Professions has provided a resource guide <https://docs.google.com/document/d/1O9XBMSMXxrdtN102xn7gQmJqew_dbFatY0d1CJL…> to help members implement the recommendations of this report.

July 24, 2025