On this episode of BeyondPhrenology, I speak with Dr. Gabriele Scheler (Carl Correns Foundation for Mathematical Biology) about the state of AI and neuroscience—past the hype, and closer to their limits.

We begin by revisiting what AI once meant: symbolic systems, logic, early neural networks, and the long-standing divide between learning and reasoning. Against that backdrop, we examine the current moment, where large language models dominate the conversation but remain, in many ways, underwhelming relative to the broader ambitions of artificial intelligence.

The discussion then turns to neuroscience, where despite decades of experimental progress, a central problem remains unresolved: the absence of integrated, functional models of cognition. We explore the consequences of a synapse-centric view, the limits of current theoretical approaches, and why accumulating more data—without perspective—fails to move the field forward. Along the way, we touch on issues that rarely make it into official narratives: the role of funding structures, the drift toward mediocrity, and the persistence of poorly framed questions.

From there, we consider an alternative direction. Dr. Scheler outlines a neuron-centric, function-driven approach to modeling the brain—one that emphasizes modularity, one-shot learning, internal inference, and decision-making as a unifying principle across cognition and emotion. Framed through evolution, this perspective highlights how biological systems bridge scales of structure and function in ways current models largely fail to capture.

The episode closes by reflecting on what this means for the future: not just for AI, but for science itself—how research cultures shift, why fascination alone is not enough, and what it would take to build models that are not just complex, but actually explanatory.
Watch at: www.youtube.com

Tracing the historical dynamics of science can reveal how scientific knowledge emerges and evolves over time. Because scientific knowledge is embedded in increasingly complex systems, comprising shifting relationships among people, the organisms and matter they study, technology, data, publications, and the concepts they utilize, scholars are looking beyond traditional historiographical methods towards quantitative and computational tools. Big data, network analysis, and machine learning enhance the scale and speed of analysis, but these methods often ignore or erase the critical roles that context (like time period, geography, and discipline) and different types of data (like image and audio data) play in the development of new knowledge. In this talk, I present context- and data-sensitive computational methods that extend efforts to model the evolution of science as a complex system. These methods reveal when new knowledge emerges and how the features of old scientific information constrain features of new scientific knowledge.

Read the full article at: www.mivideo.it.umich.edu

Theoretical computer science has over the years sought more and more refined answers to the question of which mathematical truths are knowable by finite beings like ourselves, bounded in time and space and subject to physical laws. I'll tell a story that starts with Godel's Incompleteness Theorem and Turing's discovery of uncomputability. I'll then introduce the spectacular Busy Beaver function, which grows faster than any computable function. Work by me and Yedidia, along with recent improvements by O'Rear, Riebel, and others, has shown that the value of BB(549) is independent of the axioms of set theory; on the other end, an international collaboration proved last year that BB(5) = 47,176,870. I'll speculate on whether BB(6) will ever be known, by us or our AI successors. I'll next discuss the P!=NP conjecture and what it does and doesn't mean for the limits of machine intelligence. As my own specialty is quantum computing, I'll summarize what we know about how scalable quantum computers, assuming we get them, will expand the boundary of what's mathematically knowable. I'll end by talking about hypothetical models even beyond quantum computers, which might expand the boundary of knowability still further, if one is able (for example) to jump into a black hole, create a closed timelike curve, or project oneself onto the holographic boundary of the universe.

Watch at: www.youtube.com

Venki Ramakrishnan 30th Ulam Lecture Night 2 The knowledge of aging and death has driven human culture, including our religions, ever since we became aware of our mortality. For much of our existence there was not much we could do about it. But over the past few decades, biology has made major advances in our understanding of the causes of aging, opening for the first time the possibility of intervening in the process. At the same time, the combination of longer lives and reduced fertility rates means that many societies are faced with an aging population. This has led to large investments in aging research from governments and private industry funded largely by tech billionaires, resulting in both real advances and a large amount of hype. In this talk, Venki Ramakrishnan will discuss some of the key findings about why and how we age and die and prospects for the future. He will also explore the possible consequences of societies with extremely long-lived populations.

Watch at: www.youtube.com

Computing is not just a branch of engineering. It touches on language and philosophy, the nature of life, and how we think. It is a liberal art. And as a result, we need a humanistic approach to computing. By understanding the history of technology more deeply, as well as how it connects to so many disciplines, we will be better positioned to make this approach the default in how we relate to computing.

Sam Arbesman's new book "The Magic of Code" explores this. But, he is also interested in more broadly developing what he is calling the Humanistic Computation Project, which is aimed at creating a living syllabus, a community, and a framework for these ideas.

Watch"Humanistic Computation and The Magic of Code" Dr. Sam Arbesman at: www.youtube.com

In the mid-20th century, Alan Turing and John von Neumann developed the theoretical underpinnings of computer science, neuroscience, and AI. They also founded the field of theoretical biology, showing how living systems must necessarily be computational in order to grow, heal, and reproduce. Recent experiments by Blaise Agüera y Arcas’ team at Google have drawn new connections between theoretical biology and computer science, showing how “digital life” can evolve in a purely random universe. Such artificial life doesn’t evolve the way Darwinian evolutionary theory usually presumes, through random mutation and selection, but rather through symbiogenesis, wherein small replicating entities merge into progressively bigger ones. This may be the creative engine behind biological evolution too. In this lecture, Agüera y Arcas will describe how symbiosis explains both life’s origins and its increasing complexity. He’ll also draw connections to social intelligence theories, which suggest that similar symbioses have powered intelligence explosions in humanity’s lineage and those of other big-brained species. Finally, he’ll argue that both modern human intelligence and AI are best understood through this symbiotic lens.

Watch at: www.youtube.com

Venki Ramakrishnan 30th Ulam Lecture Night 1 Ramakrishnan will provide a history of molecular visualization, as well as take us through his work at the MRC Laboratory of Molecular Biology in Cambridge, England, where his team determined the atomic structure of the 30S ribosomal subunit and its complexes with ligands and antibiotics. Everyone is familiar with DNA, but by itself, DNA is just an inert blueprint for life. It is the ribosome — an enormous molecular machine made up of a million atoms — that makes DNA come to life, turning our genetic code into proteins and therefore into us. He will talk about the ribosome (the "Gene Machine"), and how his team learned about its structure. He will also share some recent developments, including the development of cryoEM — a powerful technique used to determine the structure of three-dimensional structure of biological molecules at near-atomic resolution.

Watch at: www.youtube.com

Hiroki discussed the relationship between idea generation and social norms, explaining that new ideas can push local social norms in new directions.

Watch at: www.youtube.com

AI for Humans

Alex Pentland. Stanford HAI Fellow and MIT Toshiba Professor

Current AI is designed as a rough emulation of human intelligence. If instead we designed AI to complement human intelligence, we can achieve much more useful performance. I will show examples from finance, science, health, patents, and policy.

Watch at: www.youtube.com