Extopic, The first launch of a new type of chip that uses probabilistic chips, has produced its first working hardware and evidence that advanced processes can tackle tasks useful in artificial intelligence and scientific research.
Startup Chips work in a fundamentally different way than chips from Nvidia, AMD, and others, and they promise that they are thousands of times more powerful when they work. With AI companies pouring billions of dollars into building data centers, a completely new approach could provide a cost-effective alternative to the massive array of conventional chips.
Extopic calls its processors procermodynamic samplers, or tsus, as opposed to central processing units (CPUS) or graphics processing units (GPUS). Tsus uses silicon components to move thermodynamic Electron dynamics, shaping them into models of possibilities for various complex systems, such as weather, or AI models capable of generating images, text, or videos.
The first additional operating chip has been shared with several partners, including Frontier AI Labs, climate modeling startups, and several government representatives. (Extopic declined to provide names.)
“This allows all kinds of developers to kick the tires, who find disagreements in the world of technology like polosophy with opposites called Beff Jeff and E / Acc before founding the startup. Verdon and his cofounder, trever mccourt, working as CTO, who previously worked with quantum computing at Google before following their computer path.
One of those now testing the new hardware is Johan Mathe, CEO of Atmo, a startup that uses AI models to make predictions with higher resolution than is possible. Customers include the Department of Defense. Mathe says that the extopic chips should make it possible to calculate the odds of different weather conditions very well.
Extopic also releases software called trhml that makes it possible to synchronize the behavior of the extopic chip to the GPU. Mathe has used this software and the real chip. “I was able to run a few p-bits and see that they behave as they should,” Mathe said.
The company’s hardware, called XTR-0, consists of a Field-Programmess Grate Array (FPGA) chip, which can be reconfigured with different functions, combined with its first chip, the X-0, each of which contains a few qugits.
Instead of the usual bits that correspond to 1 or 0, the new chip includes plumbolistic bits, or p-bits, that matter is uncertain. Although limited in scale, the new chip shows the power of the company’s new approach.
“We have a machine learning primitive that is much more efficient than a matrix of iterations,” McCourt said. “The question is, How do you build something on the scale of Chatgpt or Midjourney.”
