Stanford Institute for Human-Centered Artificial Intelligence (HAI)’s Post

Today, I completed my summer AI Ethics course at the University of California, Berkeley, and it has completely transformed my way of thinking. While many of my peers are diving into traditional data science, I chose to explore the ethical dimensions of AI. This experience has deepened my understanding of the impact technology has on society, and I'm excited to apply these insights as I continue my journey in data science. Looking forward to what's next.

This short essay examines AI’s algorithmic bias and its ethical implications.

If you're concerned about the ethical practices of high-tech companies, especially regarding AI and its impact on our digital lives, this program is a must-see! Don't miss out on the opportunity to learn more about democratic regulation in the tech industry. #AIethics #techregulation

"...or when developers get carried away with doing something “cool” with AI, and shifting away from the core purpose of problem-solving for end users" ...or when tax-payer funded professors get carried away with doing something "cool" with AI and shifting away from the core-purpose of research and education.

// The Only Possible Objects of Trust Are Either the Institutions or People Responsible for Technology // Interesting design anthropological analysis of how trust and trustworthiness concepts are articulated and performed by AI software practitioners. I especially appreciated the literature review describing the differences in how trust is perceived in different fields. "while agendas for trustworthy AI superficially engage shared terminology, they in fact frequently represent diverse and contested understandings of what trust, trustworthiness and ethics entail, and different alignments with the relations of power and capital." "these suggest that trust between humans and machines is impossible and therefore argue that the only possible objects of trust are either the institutions or people responsible for technology." "instead, design anthropology turns around the relationship between trust and those other agents that might be related to it. It begins with the premise that if we understand trust as an underpinning element (and experiential possibility) of the circumstances of human life, we can then ask how AI and other technologies and things might become part of those circumstances of trust." "we must connect the way that trustworthy AI development is taught to software engineers more closely /to the ways that trustworthy AI is experienced, expected and anticipated by those who use it in everyday life/." Data & Society Research Institute Sareeta Amrute Cc. Amanda Andres Amy Ko

Much before all of this started, in my mind, there was the thought that we could not compete with the "machines." Imagine the program or LLM that has the source of all available information for certain topics, has the skills to compose work based on the rules of the best textbooks, knows all the statistical methods and rules, and has perfect language knowledge. This model could generate the most comprehensive reviews and analyses in seconds. How can my knowledge compete with this or anyone's knowledge? And this will happen, sooner or later. With all the troubles in the scientific community that already exist, will this change be the death sentence to science or a new beginning for better science that we are not aware of yet? https://lnkd.in/dq57dSKi

'I've always felt that law has a special role to play in making sure that we can use technology for good.' Discover the research of Sandra Wachter, Professor of Technology and Regulation at the Oxford Internet Institute, University of Oxford - including her key areas of focus around #AI and her predictions for the future of her field ⬇️ 

At Data School, we do not want to merely study the digital society, we want to take part in shaping it. The societal impact of our work is an important result of our approach to research. But there is also an epistemic angle to it: when working with societal partners, we have the opportunity to study AI and data practices up close. We are where technological change manifests. [Iris Muis and I refused to have our pictures taken for the interview; this team photo is a compromise, it does not show former members, our university colleagues or our external partners; what we do is #teamscience, it's inter- and even transdisciplinary, and it rocks!] https://lnkd.in/etBNuBgs

Excited to share my latest research paper! It is a condensed version of my M.A. thesis and represents the final research output of my work as a Fulbright and DAAD scholar in the U.S. and the U.K. It proposes a (geo)political risk taxonomy of AI, identifying 12 risks distributed across 4 categories: (1) Geopolitical Pressures, (2) Malicious Usage, (3) Environmental, Social, and Ethical Risks, and (4) Privacy and Trust Violations.  Incorporating a regulatory side, the paper conducts a policy assessment of the EU AI Act. The landmark regulation has the potential to have a positive top-down impact concerning AI risk reduction but needs regulatory adjustments to mitigate risks more comprehensively. Regulatory exceptions for open-source models, excessively high parameters for the classification of GPAI models as a systemic risk, and the exclusion of systems designed exclusively for military purposes from the regulation’s obligations leave room for future action.