TEDxMIT: Ionics for Energy-Efficient AI Hardware (Bilge Yildiz)

Energy demand and CO2 emissions of AI is increasing unsustainably. Revolutionary advances are needed from both algorithms and hardware technologies to improve the energy efficiency of computing by more than one million-fold. Physical neural networks combine processing and memory in one unit, and can enable significant gains in energy efficiency of computing for AI. This talk introduces ionic synaptic devices to function as programmable node weights in such physical neural networks. Inspired from the biological synapse, these devices work with shuttling of ions that lead to conductance changes at the synapse, and are able to do this with extremely low energy (lower than the biological synapses), at very fast speeds and reliably. As a result, these devices can enable both near-term energy-efficient deep neural networks that can be integrated into next generation AI chips, as well as longer term and more bio-realistic networks such as spiking neural networks and emulating neuroscience-deduced learning rules for more capable AI hardware.

TEDxMIT: Neutrons for Networks, Powering AI Sustainably with Nuclear Energy (Jacopo Buongiorno)

AI, Energy, Environment, Technology, Nuclear energy Jacopo Buongiorno is the TEPCO Professor of Nuclear Science and Engineering at the Massachusetts Institute of Technology (MIT), the Director of the Center for Advanced Nuclear Energy Systems (CANES), and the Director of Science and Technology of the MIT Nuclear Reactor Laboratory. He has published over 100 journal articles in the areas of reactor safety and design, two-phase flow and heat transfer, and nanofluid technology.

No Fuel, No Party Symposium

To throw a nuclear party, you need the right fuel. That message echoed throughout MIT's October symposium, where experts from across the nuclear energy landscape gathered to confront a rapidly evolving sector driven by technological advances, updated regulation, shifting geopolitics, and pressing questions about fuel supply.

Nuclear Everywhere?" Symposium

Presenters and panelists for the "Nuclear Everywhere?" Symposium offered videos from their institutions, which we assembled for display during the symposium. The symposium ran from March 28th to March 29th, 2023 on the MIT campus.

Nuclear Energy: A New Beginning

The MIT Future of Nuclear Energy in a CarbonConstrained World study is the eighth in the MIT Energy Initiative’s “Future of” series, which aims to shed light on a range of complex and important issues involving energy and the environment.

Heat-Seeking Studies

Months of painstaking setup and delicate experimentation paid off recently for Artyom Kossolapov with a thrilling moment of discovery. It came while the nuclear science and engineering student was conducting research for his master's degree, as he observed a metallic device immersed in water.

Floating nuclear plants could withstand earthquakes and tsunamis

A new design for nuclear plants built on floating platforms, modeled after those used for offshore oil drilling, could help avoid such consequences in the future. Such floating plants would be designed to be automatically cooled by the surrounding seawater

Mastering Fluid Flow and Bubble Boil

Etienne Demarly’s efforts in generating physically realistic computer simulations may soon enable current generation reactors to operate more efficiently.

 

Gridlocked Podcast

GRIDLOCKED is a mixed-format docuseries podcast covering key solutions to many of the world's fundamental challenges. Each season focuses on a major intersection of obstacles in the way of human progress, promoting an improved public discourse as we look to move forward.