Computational Dyson Spheres and AI
Speaker
Jason Wright and Joseph Bates
Penn State and AI for Humanity Foundation
Host
Gerry Sussman
CSAIL
Abstract:
There is a longstanding expectation that if spacefaring technological species exist elsewhere in the galaxy, that they will build vast computers in space that harvest starlight to perform calculations on a massive scale. Jason will discuss ongoing efforts to search for such "computational Dyson spheres", and some of the outstanding issues surrounding their nature, including the fundamental thermodynamic limits that constrain computation in space, and how the practicalities of orbital dynamics present novel problems of high performance computer architecture. Joe will offer thoughts on how these machines might support large scale AI and related algorithms.
We also will invite discussion on many open problems in the field, such as: How can nodes synchronize and communicate when their physical arrangement may change on a millisecond time scale? How can such systems manage latencies set by vast physical distances? What is the optimal configuration of nodes in orbit as constrained by the laws of gravity and the need for collision avoidance? How might advanced AI systems take advantage of the extraordinary computing power of this hardware?
We are planning a 2025 interdisciplinary workshop on these topics, and hope this talk will attract interested experts from computer science, physics, economics, evolution, and other relevant fields.
Bios:
Jason Wright is a Professor of Astronomy and Astrophysics at Penn State and the director of the Penn State Extraterrestrial Intelligence Center. He is a stellar and exoplanetary astrophysicist specializing in the detection and characterization of planets orbiting other stars, and stellar magnetic and convective activity. He has pioneered novel observational campaigns in radio SETI and is an expert in the search for Dyson spheres, including their thermodynamics and observational signatures. He was the recipient of the 2019 Drake Award for his achievements in SETI and exoplanets, was selected by NASA to chair the 2018 NASA Technosignatures Workshop that marked NASA's reentry to the field, and is the author of a forthcoming textbook on SETI.
Joseph Bates is the founder of Singular Computing, which has developed approximate computing technology widely used in modern deep learning hardware. He also leads the AI for Humanity Foundation, which is working to provide universities with billions of cores of computing, to enable advanced AI in the open. His technical activities have been varied, including automated mathematics and type theory, hardware synthesis and simulation tools, computational drama, approximate computing, and AI. He was a professor and scientist for 20 years at Carnegie Mellon, the MIT AI and Media Labs, Cornell, where he received his PhD, and Johns Hopkins, which he entered at age 13.
This talk will also be streamed over Zoom: https://mit.zoom.us/j/99642185015.
There is a longstanding expectation that if spacefaring technological species exist elsewhere in the galaxy, that they will build vast computers in space that harvest starlight to perform calculations on a massive scale. Jason will discuss ongoing efforts to search for such "computational Dyson spheres", and some of the outstanding issues surrounding their nature, including the fundamental thermodynamic limits that constrain computation in space, and how the practicalities of orbital dynamics present novel problems of high performance computer architecture. Joe will offer thoughts on how these machines might support large scale AI and related algorithms.
We also will invite discussion on many open problems in the field, such as: How can nodes synchronize and communicate when their physical arrangement may change on a millisecond time scale? How can such systems manage latencies set by vast physical distances? What is the optimal configuration of nodes in orbit as constrained by the laws of gravity and the need for collision avoidance? How might advanced AI systems take advantage of the extraordinary computing power of this hardware?
We are planning a 2025 interdisciplinary workshop on these topics, and hope this talk will attract interested experts from computer science, physics, economics, evolution, and other relevant fields.
Bios:
Jason Wright is a Professor of Astronomy and Astrophysics at Penn State and the director of the Penn State Extraterrestrial Intelligence Center. He is a stellar and exoplanetary astrophysicist specializing in the detection and characterization of planets orbiting other stars, and stellar magnetic and convective activity. He has pioneered novel observational campaigns in radio SETI and is an expert in the search for Dyson spheres, including their thermodynamics and observational signatures. He was the recipient of the 2019 Drake Award for his achievements in SETI and exoplanets, was selected by NASA to chair the 2018 NASA Technosignatures Workshop that marked NASA's reentry to the field, and is the author of a forthcoming textbook on SETI.
Joseph Bates is the founder of Singular Computing, which has developed approximate computing technology widely used in modern deep learning hardware. He also leads the AI for Humanity Foundation, which is working to provide universities with billions of cores of computing, to enable advanced AI in the open. His technical activities have been varied, including automated mathematics and type theory, hardware synthesis and simulation tools, computational drama, approximate computing, and AI. He was a professor and scientist for 20 years at Carnegie Mellon, the MIT AI and Media Labs, Cornell, where he received his PhD, and Johns Hopkins, which he entered at age 13.
This talk will also be streamed over Zoom: https://mit.zoom.us/j/99642185015.