Georgia Millman-Perlah’s Post

🌏 If you work with research in the AI space or tech in general and care about inclusion here is a must read: “Surveys Considered Harmful? Reflecting on the Use of Surveys in AI Research, Development, and Governance” (link in comments) One of the points in the paper that stood most out to me is the risk of surveys being designed from a Western centric perspective skewing the research results and hereby in the end the technology being designed. Thanks to Mohammad Tahaei, Ph.D. - one of the authors - for sharing the article with me.

Looking forward to GW’s Imagining the Future with AI on April 2, 2024. A day of #AI seminars, discussion and networking with executives and leaders. An opportunity to gain their insights on the topic and how their organization is embracing and managing the #futureofwork. I’m curious to learn more about plans to manage #employeeengagement, #upskilling, #reskilling and #managerdevelopment.

The researcher, Jose Fernando Jimenez from Université Savoie Mont Blanc, successfully presented today at the SOHOMA 2024 Workshop, as part of the Ethics4.0 project. The topic focused on "Ethical Issues in the Use of Generative AI in Performance Management." A great discussion on the industrial practical views and key ethical challenges! Congratulations for authors, Damien Trentesaux Lamia Berrah and GUERRE-CHALEY Frederic for this interesting study. SOHOMA Laboratoire SYMME LAMIH UMR CNRS 8201 Main message: Companies must find a balance between leveraging Gen AI and preserving human capabilities, ensuring sustainable performance in the long term of industrial operations

Incredibly excited to share our new paper, What is Civic Participation in Artificial Intelligence? by Renee Sieber, myself, Suthee Sangiambut and Abigail Adu-Daako This paper has been in the works for *years*! From our analysis, we derive five themes from authors’ work on participation in AI:  (1) participation as a natural byproduct of automating government (2) participation facilitated through the medium of AI (3) participation in AI as quantification (4) participation as a technocracy of trust, and (5) participation as meaningful "With few exceptions, authors seemed not to challenge the status quo nor diminish the authority of the experts. Authors focused on the processual without the influence and AI aided in that process orientation. We conclude that the future of public participation in AI requires careful attention to become meaningful including recognition of neoliberal intent and power differentials." It's open access in the Journal of Environment and Planning B: Urban Analytics and City Science https://lnkd.in/eY22PKiR

On February 14th at 15:00 CET I will participate in a panel discussion with four PhD fellows about the implications of Generative AI. We will discuss along the lines of sustainability, legal implications, accountability, alignment and more!

In our recent op-ed for Scientific American, Franck Marchis and I explore the idea of using LLMs to connect with extraterrestrial civilizations, reminiscent of the Voyager Golden Record. A project like this raises significant technical, societal, and philosophical considerations that need to be defined and addressed collectively. E.g., how should we send it? What interface should we adopt? How do we encapsulate the essence of humanity within a single LLM? What if an extraterrestrial civilization misuses this information? Our goal with this piece is to spark a conversation on the potential of AI in the realm of METI/SETI and to explore the exciting possibilities this technology offers for interstellar communication. https://lnkd.in/eX_GTf24

Reflecting on the discussions in this MIT article, it's crucial to ensure AI enhances human connections. Maintaining a human-centric approach in technology design is key. Aligning economic incentives with ethical standards ensures AI enriches our lives. Let's focus on creating environments that foster meaningful human connections. How can we effectively balance technology with human interaction?

The article "A Teleological Approach to Information Systems Design" is finally available for open access in the journal "Minds and Machines". As I anticipated in a previous post, we propose an approach that can be used to support the analysis and re-design of information systems based on a well-known information theory, namely teleosemantics. Our objective has been also to emphasize the centrality of the responsibility and awareness of the agents involved in the design and use of artifacts that are increasingly difficult to categorize and increasingly characterized by general and ill-defined purposes. Although it has taken us two years, we see this work as a first step, toward many directions. We also see it naturally in synergy with significant studies advocating for regulation and recognizing the importance of non-neutrality in AI development and possible related risks, such as those by Luciano Floridi and Mariarosaria Taddeo, whose work has greatly inspired us. Here is the link to the full article: https://lnkd.in/dqpjM5eB We hope this contribution will be of interest. We are looking forward to your feedback!

With #AI becoming more and more human-like (in the sense of imitating or replicating human behavior), what is the level of trust we can have in these systems? Even if over time, #hallucinations can be further reduced or even eliminated completely (then again, they're part of human behavior as well), there's other risk that will stay: humans "tend to attribute a great deal of authority and trustworthiness to machines" and such misplaced faith can also cause problems when "AI systems are used for research" (https://lnkd.in/djGTXq_x). The danger is that AI tools are anthropomorphized as something like superhumans which they are clearly not. Another risk is AI monoculture along 2 dimensions. The "monoculture of knowing" is based on the fact that AI systems are good in the domains they've been trained in which limits the overall set of questions they can answer, leaving out other aspects they simply don't know. The "monoculture of the knower" refers to AI tools having a specific standpoint based on their training, likely eliminating different perspectives. Both are different if there many kinds of (human) minds working on a problem which makes it "more likely to spot false assumptions or missed opportunities" (depending on the group of people of course). #genai Scientific American 17 https://lnkd.in/d5kU3jMi

📢 Announcing the Second Edition of the AI Action Summit Newsletter (in French and English)! I'm excited to share the latest insights and developments in the preparation for the AI Action Summit. All links are in the comments. 1. How to Participate: 📝 Expert consultation is open until November 15 on the Summit's website 💡 Respond to the call for projects by October 7, 2024, managed by the Paris Peace Forum. 2. In this edition, I explore methodologies for international convergence in AI governance, offering pathways to balance a reliance on existing scientific evidence with the need for political innovation in AI policy. 👉 Read the full newsletter here: aias.news