Cameron Jones and Benjamin Bergen (31 Mar 2025). Large Language Models Pass the Turing Test. arXiv:2503.23674 This paper seems worthy of contemplation, and clearly celebration for this team from UC San Diego for demonstrating the first empirical evidence that any artificial system passes a standard three-party Turing Test. Time to up the bar. Can the AI nae nae? Determining an intelligent system requires deep review and contemplation of what constitutes intelligence and how this can be observed and verified.

Friston et al (2024). Federated inference and belief sharing. Neuroscience and Behavioral Reviews (Vol 156, Jan). I’ve been thinking a lot lately about Bayesian inference, mostly in the context of systems of devotion over long time scales, and how we might disentangle true collective will from egregious manipulation of such will, which is something that requires ethical consideration by all who endeavor to preserve and advance human dignity. The paper begins with three clearly stated highlights addressing communication as an emergent property of conspecifics, how “evidence-maximizing” processes can explain the emergence of language and belief transmission across generations, and how leveraging Bayesian terms can ground these concepts.

Lee ED, Flack JC, Krakauer DC. 2024 Constructing stabliity: optimal learning in noisy ecological niches. Proc. R. Soc. B 291: 20241606 One of my research questions is how do acquatic creatures, including but not limited to whales, learn and form representational understanding of their surroundings in dynamic and sensory rich contexts (the ever changing ocean). This paper frames two ancillary questions - how quickly should an organism respond to a changing environment and what is the optimal trade-off between remembering and forgetting?

Centre for the Governance of AI Annual Report (2023) I’m realizing my motivation to read and journal so much about AI safety is as equally motivated by wanting to be adaptive to emerging technologies as an engineering manager as well as being a concerned citizen. There can be this tacit belief that there is a fleet of really smart people out there who are asking all the right questions, and there definitely is, but there’s been such a long history of engineering leadership spaces that were not representive of the populace at large that I find it particularly compelling to look around documents like this annual report and ask myself if I would approach the subject matter differently.

Persistent Barcodes for Shapes by Gunnar Carlsson, Afra Zomorodian, Anne Collins and Leonides Guibas (Eurographics Symposium on Geometry Processing, 2004). This is a meaty paper with an express purpose of introducing us to topology concepts that help us classify shapes based on their global connectedness. The authors aim to leverage local and global information to characterize shapes. A core concept introduced is that of the filtered tangent complex, which is a family of spaces paramterized by curvature.

Human-AI Decision Systems by Alex (Sandy) Pentland, Matthew Daggett, and Michae Hurley. This paper outlines a proposal for a high-performance human-AI decision system, centering on how trust in a human-AI decision system and its performance could be ensured in a commercial enterprise context. The paper also explores how multi-domain task teams could be created to oversee operations. Along the way the authors outline some of the key challenges, including how such decision systems can be built in the context of legacy architecture.

Physics-informed neural networks: A deep learaning framework for solving forward and inverse problems involving nonlinear partial differential equations. M. Raissi, P. Perdikaris, G.E. Karniadakis. Journal of Computational Physics 378 (2019) 686-707 The problem: “There exists a vast amount of prior knowledge that is currently not being utilized in modern machine learning practice. Let it be the principled physical laws that govern the time-dependent dynamics of a system, or some empirically validated rules of other domain expertise, this prior information can act as a regularization agent that constrains the space of admissible solutions to a manageable size”.

Measuring algorithmically infused socities. Nature. 30 June 2021. Wow wow wow. Such a great perspective paper centering some of our collective concerns that while algorithmic tuning of processes affords many efficiencies, entrenched biases and inequalties can also get exacerbated. We’ve been doing this long enough for us to start measuring some of the structural developments in algorithmically-infused societies. The authors zoom on three challenges to measuring what’s going on in such societies — not enough measurment coverage of the problem space, a host of problems that crop up with mis-measurement, and the limited frameworks for us to even grok this problem set at all.

AI and It’s Implications for Income Distribution and Unemployment (2017) by Anton Korienk and Joseph E. Stiglitz. Background paper for NBER conference “The Economics of AI”. Technological Progress and Welfare: A Taxonomy First Best First-best scenarios are those in which we assume all markets are perfect. I’m not sure what utility there is in discussing such impossible states as these, but the authors seem to position them as a baseline ideal we can compare all other, viable, scenarios.

AI Governance: A Research Agenda (2018) by Allan Dafoe. Centre for the Governabnce AI, Future of Humanity Institute, University of Oxford. Eye on the prize ya’ll This review is filled with measured calls to ideation sweeping across a host of AI ethical concerns, yet the sentiment that rings out as most primary is the call to vision stated thus, “AI ideal governance aspires to envision, blueprint, and advance ideal institutional solutions for humanity’s AI governance challenges.