A new method could increase the training efficiency of large language models: By leveraging idle computing time, it can double the speed of model training while preserving accuracy. “Our goal was to turn this idle time into speedup,” Qinghao Hu says.
TPP graduate student Strahinja Janjusevic brings an international perspective and US Naval Academy education to his work in maritime cybersecurity. He aims to strengthen ties between the US and its allies on national security, AI, and cybersecurity.
By exposing biases, moods, personalities, and abstract concepts hidden in large language models, a new method developed in part by an MIT mathematician could root out vulnerabilities and improve AI safety and performance.
Via @lincoln_laboratory on Instagram: ”Air Force Reserve Officers’ Training Corps (AFROTC) cadets from around New England traveled to Lincoln Laboratory's Lincoln Space Surveillance Complex to learn more about space situational awareness and radar.”
MechE student Kiyoko “Kik” Hayano’s path — from Wyoming to MIT to Arkansas via D-Lab — reflects a common trajectory of U.S. innovation: talent emerging from rural places, developing on the nation’s campuses, and returning know-how to its heartland.
In a world without MIT, radar wouldn’t have been available to help win World War II. We might not have email, CT scans, time-release drugs, photolithography, or GPS. And we’d lose over 30,000 companies, employing millions of people. Can you imagine?
Since its founding, MIT has been key to helping American science and innovation lead the world. Discoveries that begin here generate jobs and power the economy — and what we create today builds a better tomorrow for all of us.