Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
extroverted, A startup is developing an exotic new type of computer chip which handles possible bits, has developed its first working hardware, and with evidence that more advanced systems can handle useful tasks Artificial intelligence and scientific research.
The startup’s chips work in a fundamentally different way than chips from Nvidia, AMD and others, and they promise to be a thousand times more energy-efficient when scaled up. With AI companies pouring in Billions of dollars in data center constructionAn entirely new approach could offer a much less expensive alternative to a vast array of conventional chips.
Extropic calls its processors Thermodynamic Sampling Units, or TSUs, as opposed to Central Processing Units (CPUs) or Graphics Processing Units (GPUs). TSUs use silicon material to exploit thermodynamic electron fluctuations, modeling the potential of various complex systems such as weather, or AI models capable of generating images, text or video.
The first working Xtropic chip has now been shared with a handful of partners, including Frontier AI Labs, startups working on weather modeling, and several government representatives. (Extropic declined to provide names.)
“It allows all kinds of developers to kick the tires,” said Extropic CEO Guillaume Verdon, who has earned a reputation in the tech world. Colorful and sometimes controversial Call people online Based on beef jezos and a new techno philosophy known as Effective Acceleration or e/acc before establishing startup. Verdon and his co-founder, Trevor McCourt, who is Extropic’s CTO, previously worked Quantum computing Before following their innovative computing methods at Google.
One of those now testing the new hardware is Johan Math, CEO of Atmo, a startup using AI models that can make predictions with higher resolution than otherwise possible. Its customers include the Department of Defense. Math says that Extropic’s chips should make it possible to more efficiently calculate the odds of different weather conditions.
Xtropic is also releasing software called TRHML that makes it possible to simulate the behavior of the Xtropic chip on a GPU. Math used this software as well as the actual chip. “I’ve been able to run a few P-bits and see that they behave the way they do,” says Math.
The company’s hardware, called the XTR-0, consists of a field-programmable gate array (FPGA) chip, which can be reconfigured for different tasks, combined with two of its first possible chips, the X-0, which each have a few p-bits.
Instead of the conventional bits corresponding to a 1 or 0, the new chip features probabilistic bits or p-bits that model uncertainty. Although limited in scale, the new chip demonstrates the potential of the company’s new approach.
“We have a machine-learning primitive that is much more efficient than matrix multiplication,” McCourt said. “The question is, how do you build something at the scale of ChatGPT or Midjourney.”
