To review this article, select My Profile and then View Saved Stories.
Steven Levy
We’ve all been amazed at how well chatbots seem to perceive the world. What if they were genuinely connected to the real world?What if the knowledge established in the chat interface was the physical truth itself, captured in real-time by interpreting the knowledge of millions of sensors spread around the world?That’s the concept of Archetype AI, an ambitious startup launching today. As co-founder and CEO Ivan Poupyrev says, “Think ChatGPT, but for the physical truth. “
Yes, they know about Apple’s much-lamented wearable device that was killed by Steve Jobs in 1997, and no, they don’t seem to care. The new Newton is designed to process sensor data of all kinds and answer questions or provide graphs or even computer code to inform what’s happening in the world. For Poupyev himself, Archetype realizes a long-held confidence that the virtual world can provide a way to interact deeply with the physical world. Though I didn’t know it at the time, it was there: The Soviet-born Poupyrev’s fascination with manipulating the world through generation was sparked in part through my e-book Hackers, which had given his father a stopover in China. “The idea that things can simply be hacked and their nature can simply be replaced by the invention of new technologies animated me for the rest of my life,” he tweeted about the ebook in 2020. He also perfected his English by reading and re-reading the e-book. You are welcome.
Poupyev’s adventure after leaving the Soviet Union and becoming a computer scientist included stints at Sony, Disney and, until last March, Google’s ATAP division. It was there that he led a team working on Soli, a task that involved integrating small radars into wearable devices to allow them to respond to a person’s gestures and movements. The demonstrations were impressive, but this technique had limitations. “Analyzing the sensors was difficult. You read them by hand,” he said. When LLMs emerged, Poupyrev and his colleagues learned that, with modifications, they could make sensor information more robust by allowing humans to seamlessly explore and monitor information collected over vast expanses of time and space. space. Instead of a great language model, it would be a great habit model. “We were excited to see how they could work with data from the physical world in real time,” he says. They were particularly excited to do it outside of Google, without the limitations of working within a giant organization. In March of last year, Poupyrev and eventually four others left to launch Archetype, now funded through a $13 million seed investment round.
Nick Gillian, Brandon Barbello, Ivan Poupyrev, Jaime Lien and Leonardo Giusti leave Group R
Archetype’s AI model, Newton, collects knowledge from other sensors, combines it, and converts it into plain language descriptions of what’s in the physical world.
In Your Inbox: Introducing Politics Lab, Your Guide to Election Season
Do you think Google’s “Incognito Mode” protects your privacy?Think again
Reporting Sexual Harassment and Assault in Antarctica
The earth will dine with cicadas.
Update your Mac? Here’s what you’ll spend your money on
Jenna Scatena
Reece Rogers
Kate Knibbs
Leif Wenar
Steven Levy
Luisa Matsakis
knight
Carolina Haskins
More from WIRING
Reviews & Guides
© 2024 Condé Nast. All rights reserved. WIRED would possibly earn a portion of sales of products purchased through our site as part of our partner component partnerships with retailers. Curtains on this site may not be reproduced, distributed, transmitted, cached, or otherwise used unless with the prior written permission of Condé Nast. Ad Choices