Alexander Wissner-Gross
| Alexander D. Wissner-Gross | |
|---|---|
| Occupations | Physicist; Computer scientist |
| Fields | Physics; Computer science; Artificial intelligence |
| Name | Alexander Wissner-Gross |
| Research topics | Entropy; Agency; Machine intelligence |
| Notable ideas | Causal entropic forces |
| Known for | Causal entropic forces; Adaptive intelligence |
| Notable publication | "Causal Entropic Forces" (2013) |
| Wikidata | Q23727955 |
Alexander D. Wissner-Gross is an American physicist, computer scientist, and entrepreneur known for unconventional ideas linking physics to computation and intelligence. A former Harvard and MIT researcher, he has proposed that intelligent behavior can emerge from basic physical principles – most notably his theory of causal entropic forces, which suggests systems act to maximize the number of available future states (or “entropy of the future”). He has also co‐founded technology companies (such as the carbon-tracking startup CO2Stats and the AI firm Gemedy, now Reified), and is a popular science presenter (his TEDx talk on an “equation for intelligence” has over a million views). Wissner-Gross’s work spans nanotechnology, machine learning, economics, and theoretical physics, and has sparked both interest and debate in artificial intelligence circles.
Early Life and Education
Wissner-Gross grew up with a passion for science. In high school he won multiple national science competition prizes – for example, a 10th-place finish in the 1999 Intel Science Talent Search and top awards at the Intel International Science Fair in 1997 and 1999 His teenage research focused on nanotechnology and programmable matter, exemplified by a project which “fuse[d] the programming of computer science with the physical world of atoms and molecules” This early interest in “physically programming matter” set the direction for much of his later work.
He entered the Massachusetts Institute of Technology (MIT) for college and distinguished himself academically. At MIT he pursued a rare triple major in physics, electrical engineering, and mathematics, graduating at the top of his class in 2003. He was even one of MIT’s Marshall Scholars (an international graduate scholarship) – one of the last students allowed to triple-major at MIT After MIT, Wissner-Gross earned a Ph.D. in physics from Harvard University in 2007. His Harvard doctorate focused on topics like programmable matter and machine learning, and was recognized with Harvard’s Hertz Doctoral Thesis Prize, an award for outstanding Ph.D. research.
Academic and Entrepreneurial Career
After completing his doctorate, Wissner-Gross pursued a dual career in academia and technology. On the academic side, he became an Institute Fellow at Harvard’s Institute for Applied Computational Science and a research affiliate at the MIT Media Lab In these roles he worked on problems at the intersection of physics, computing, and intelligence, including machine learning and emergent behaviors in complex systems. He also taught or lectured on these subjects in various seminars and workshops.
In parallel, Wissner-Gross co-founded multiple technology ventures. In 2007 he launched CO2Stats, a company that computed the carbon footprint of Internet activities. (This startup, backed by Y Combinator, famously made headlines when an early claim that a Google search emitted about 7 grams of CO₂ was questioned by Google He later founded Gemedy, Inc. (circa 2011), which built intelligent software for government and enterprise use Today he is founder and managing partner of Reified LLC, a deep-tech investment and consulting firm. Across his career to date he has earned over a hundred distinctions and awards for science and technology.
Wissner-Gross is also a prominent communicator. He has given keynote talks at conferences on artificial intelligence and on physics (for example, a 2014 keynote on the “Physics of Artificial General Intelligence”), and his TEDx lecture “A New Equation for Intelligence” (presenting his main theory) has been viewed by millions online Science outlets note that his ideas have been featured widely in media from Science, Nature and the BBC to The New York Times and Financial Times.
Causal Entropic Forces and Other Key Ideas
Wissner-Gross’s most famous contribution is the idea of causal entropic forces. In a 2013 Physical Review Letters paper co-authored with Cameron Freer, he proposed that an intelligent agent can be viewed as a physical system that acts to maximize the entropy of its future states. In simple terms, the system ‘wants’ to preserve as many possible future options as possible. Concretely, the theory defines a generalized entropic force that pushes a system into configurations that lead to the largest number of distinct future trajectories. Wisely-Gross put the idea succinctly, saying that “intelligence is a process that attempts to capture future histories”
The paper argued that this single principle could spontaneously produce intelligent behaviors in toy simulations. A striking demonstration was an inverted pendulum: a pendulum on a movable cart (a classic control problem). An ordinary pendulum will fall to the ground, reducing its future options (it can only lie limp). But the causal-entropic model “chose” to balance the pendulum upright instead, because that upright state allows far more possible future motions (it could fall left or right, etc.). In their software (called Entropica), the cart held the pendulum upright without being explicitly told to do so In other simulations, a simple block was found to “build” a tower by placing another block on it (maximizing future reach), and even multiple agents spontaneously displayed cooperative behavior according to the same entropy principle.
In the paper’s abstract, Wissner-Gross and Freer noted that such causal-entropic dynamics could give rise to “two defining behaviors of the human cognitive niche — tool use and social cooperation” In other words, they claimed that just by physics and entropy, their model produced elementary forms of learning and collaboration without programming those goals. The effect has been metaphorically summarized by an “equation for intelligence” (in a TED talk he wrote it as ), meaning roughly “a force which accelerates a system toward states of higher future entropy”. This framework ties intelligence to adaptability: systems under this drive are said to exhibit what Wissner-Gross would term an “adaptive intelligence”, continuously adjusting themselves to keep their future options open.
Aside from entropic forces, Wissner-Gross has worked on varied subjects. In physics he co-authored papers on nanostructures and materials (for example, a 2007 Physical Review E paper showed that a diamond surface could keep ice stable up to human-body temperatures In finance he wrote an imaginative 2010 paper on “relativistic statistical arbitrage” (also in Phys. Rev. E) with Freer. That work noted that as computers trade ever faster, the finite speed of light becomes a limiting factor; they even mapped out “light-speed acupuncture points” around the globe where it would be optimal (in principle) to locate trading servers to minimize communication delays He has also published on areas like reconfigurable computing, nanowire circuits, and even efficient ways to draw barcodes, reflecting a broad interest in how information and physics intersect.
Research Methodology
Wissner-Gross’s novel ideas come from mathematical modeling and simulation. To study causal entropic forces, he and colleagues built computational models in which agents could explore many hypothetical futures. In practice this meant defining a ``thermodynamic” objective: the agent computes, for each possible present action, the path entropy of states reachable over some future time horizon. It then biases its motion in the direction that maximizes this entropy. This requires a perfect or near-perfect simulation of the environment – effectively the agent must “know” all possible outcomes (up to statistical sampling). In their paper they used a Monte Carlo–like approach to estimate entropy; the simulations were necessarily limited to very simple scenarios (few objects, short time horizons).
The Mystic force arises mathematically by generalizing the usual entropic force (like a polymer ballooning to increase entropy) into a causal form that accounts for entire future paths One can say the system is guided by the gradient of the entropy of its future distribution. Because this involves huge computational branching, it is feasible only in toy models. In fact, commentators often point out that implementing this idea in practice requires an improbably accurate forward model. As one analysis noted, it “requires the agent can actually predict possible future states” and assumes “perfect simulators” – an unrealistic assumption for most real-world AI contexts. In short, the method is currently more of a thought experiment and mathematical framework than a practical engineering algorithm.
Influence and Reception
Wissner-Gross’s ideas have attracted considerable media attention and stirred debate. Popular science outlets like CNN, the BBC, Nature, and Phys.org reported on the causal-entropic theory in 2013. For example, Phys.org described the PRL paper as “ambitious” for trying to recast intelligence as fundamentally thermodynamic His work was also reported as science news (often using whimsical phrases like “intelligence from chaos”) and he became something of a public figure in tech discussions. His “equation for intelligence” talk and keynotes have been widely viewed, and several books on AI have referenced his concepts.
Many colleagues and journalists found his cross-disciplinary approach thought-provoking, even if they remained unconvinced. Some computer scientists noted parallels between his entropic maximization and existing ideas in AI (for example, the notion of “empowerment”, where an agent seeks to maximize its potential influence over future states). Regardless, Wissner-Gross’s framing — intelligence as entropy maximization — added a new perspective to long-running debates on the nature of cognition. He was invited to speak at AI workshops and even participated in discussions about AI safety, alongside figures like Elon Musk and Stephen Hawking. His entrepreneurial ventures (CO2Stats, Gemedy) and academic role further spread his influence between industry and research communities.
Critiques and Limitations
Despite the buzz, experts have been sceptical of how well causal entropic forces solve real AI problems. Critics argue that intelligence involves many specialized skills (language, abstract reasoning, vision, etc.), so one simple law is unlikely to cover them all. In 2013 The New Yorker published a critique noting that “grand, unified, one-size-fits-all solutions almost never work” and that maximizing entropy does not by itself give the specific insights needed for tasks like language understanding This article (and others) suggested that while the idea is elegant, it underestimates real-world complexity.
Technical commentators also point out practical shortcomings. The need for a perfect model means the approach scales poorly. On a discussion forum like Hacker News, AI researchers explained that the causal-entropic agent is effectively optimizing for flexibility rather than any conventional reward – a concept not well-defined without a simulator They linked the idea to known algorithms like reinforcement learning, where typically an agent optimizes expected rewards; causal entropy instead would optimize an abstract bonus for future option richness. In short, critics say this formulation still needs a clear way to translate into working code for a real robot or app, and so far it remains a theoretical construct.
Moreover, no mainstream AI system today explicitly uses causal entropic forces as its decision rule. The field of machine learning has largely advanced through data-driven statistical methods (neural networks, etc.), and the entropy-maximization approach has not seen practical use in those domains. Some regard Wissner-Gross’s proposal as an intriguing conceptual model, akin to a physics-inspired thought experiment, rather than a ready-to-use algorithm.
Legacy and Continuing Work
Wissner-Gross is still active in research and industry, so his full “legacy” is unfolding. He continues to teach, advise, and invest in startups, and to publish new ideas at the junction of physics, computation, and intelligence. His work on causal entropic forces remains a talking point in discussions about theoretical AI — it is remembered as a bold attempt to unify ideas from thermodynamics and evolution into an algorithmic form. If nothing else, it underscores a deeper lesson: intelligence may not be easily pinned down to one simple rule, but exploring such rules can yield fresh insights.
Whether the entropic framework will prove useful in future AI systems is uncertain. So far, independent researchers have tried to build on it (for example, using approximations to estimate future entropy in reinforcement learning settings), but no breakthrough applications have emerged. Still, by sparking debate about what constitutes “intelligence” and encouraging a physics viewpoint, Wissner-Gross has influenced a niche of thinkers interested in general intelligence and the physics of information. His story also illustrates the interplay of academia and entrepreneurship in modern science: many of his ideas were showcased through public talks and startups, spreading the conversation beyond ivory towers.
Selected Works
- A.D. Wissner-Gross and C.E. Freer, “Causal Entropic Forces”, Physical Review Letters 110(16): 168702 (2013).
- A.D. Wissner-Gross and C.E. Freer, “Relativistic Statistical Arbitrage”, Physical Review E 82: 056104 (2010).
- A.D. Wissner-Gross and E. Kaxiras, “Diamond Stabilization of Ice Multilayers at Human Body Temperature”, Physical Review E 76: 020501 (2007).
- A.D. Wissner-Gross and C.E. Freer, “Pattern Formation without Favored Local Interactions”, Physical Review E 69: 062902 (2004).
- A.D. Wissner-Gross, “Dielectrophoretic Reconfiguration of Nanowire Interconnects” (conference paper, 2006).
- A.D. Wissner-Gross and collaborators, “Physically Programmable Surfaces” (annual conference on programmable matter, 2007).
Timeline (key events): Intel ISEF awards (1997–99); Intel STS finalist (1999); MIT triple major (2003); Harvard Ph.D. (2007); CO2Stats founded (2007); Phys. Rev. E on arbitrage (2010); causal entropic forces paper (2013); TEDx talk published (2014).