Is the era of models getting bigger and bigger ending? “We went

Press Space to continue

Is the era of models getting bigger and bigger ending? “We went into an era the last 5 years of building bigger and bigger and bigger models.” “We went from 175 billion parameters to 10 trillion parameters right now.” “That's going to stop. There's a lot of companies right now taking that shift as we speak, abandoning what we call the scaling hypothesis of building bigger and bigger models.” — IMEC’s Steven Latré

Molly O’Shea@MollySOShea22h ago

BREAKING: “5 to 10 years from now, we're going to laugh at how old-fashioned AI was today.” Steven Latré is VP of AI at @imec_int, one of the most important semiconductor organizations most people have never heard of. “Almost every chip in the world today has been touched by imec.” The world's biggest chip companies come to imec to work on technologies years before they reach the commercial market. I sat down with Steven & Strike Capital's Adam Chambers in Monaco to talk about what the next decade of AI hardware could look like. We cover: › Why today's AI is still incredibly inefficient › What happens as transistor scaling approaches the limits of physics › Why memory is already becoming a critical constraint › The growing importance of hardware + software co-design › Why photonics could become a “killer technology” for the chips of the future

View