Listen to this story
|
AI startups are getting more and more creative in ways one could ever imagine. Mistral AI, for instance, pasted a torrent link to its model on a post on X without any context. Extropic AI, did a similar thing, but instead of a link, posted a message in Morse code about the launch of its company’s funding round.
Best is that the company says it is receiving messages from the future, where the generative AI would be omnipresent.
Extropic AI has announced a $14.1 million seed round, boasting itself as the pioneer of physics-based AI and building a full-stack paradigm of physics-based computing. The round was led by Steve Jang and Kindred Ventures, with participation from Buckley Ventures, HOF Capital, Julian Capital, Maque VC, OSS Capital, Valor Equity Partners, Weekend Fund, and many others.
The company is also backed by notable angel investors and startup founders, including Aidan Gomez, Amjad Masad, Aravind Srinivas, Chris Prucha, and Ivan Zhang along with experienced folks like Arash Ferdosi, Balaji Srinvasan, Naval Ravikant, Scott Belsky, and Tobias Lutke.
The founder and CEO, Guillaume Verdon is also leading the “effective acceleration” movement, which is rampant in Silicon Valley right now, which stands for “move fast and make things” (check out all the e/acc accounts). Apart from his original X account, Verdon also has a handle called ‘BasedBeffJezos’ which Marc Andreessen calls as the “patron saint of techno-optimism”. e/acc is basically about accelerating the pace at which AI is being built, instead of decelerating it.
Thermodynamics in AI
In an interview, which Verdon says doxxed his identity, he said that the goal of it all is simple – “We’re kind of just optimistic people that want to build a better future. It’s pretty straightforward.” For this, Extropic aims to build a computing paradigm harnessing the power of out-of-equilibrium thermodynamics to fundamentally merge generative AI with the physics world. This will help the company achieve efficiency in building generative AI in the physical world in terms of space, time, and energy.
The team comes from various backgrounds including AI and Physics, with experience from AWS, Meta, IBM, NVIDIA, Xanadu, and Google, making it perfect for the mission.
To start with, Verdon is a former quantum tech lead at Physics & AI team at Alphabet’s X, and founded Extropic in 2022 in stealth mode. He is also regarded as the founder of Google’s TensorFlow Quantum project during his PhD at University of Waterloo. He also has a background in theoretical physics and information theory, along with a Master in Quantum computing.
Trevor McCourt, the CTO, who is originally a mechanical engineer was also one of the founding members of the TensorFlow Quantum project at University of Waterloo. He then later went on to making hardware again, eventually coming back to physical systems, and did his PhD at MIT studying the noise in computational and living systems.
Christopher Chamberland, principal architect is also from a quantum computing background and has led efforts of core quantum and roadmaps at AWS, Microsoft, and IBM Quantum.
All about open source, with a twist
Verdon highlighted in a recent post that the company believes in the power of open source and wants to bring it to the realm of computing. He added that the releases are coming soon. “Everyone should have the opportunity to build,” he said.
Extropic, which was earlier known as Qyber, has been designing microchips for running LLMs, to compete with NVIDIA. This is similar to what Sam Altman from OpenAI has been doing by investing in Rain AI to develop what it calls neuromorphic processing units (NPUs), with the same intention of increasing efficiency of chips.
Interestingly, Vernon described his company from his alternate account as working at an “AI Manhattan project” and that people would ban it if they knew what he was working on. To further highlight the importance of open source, he said in the interview, “we’ve got to make sure AI doesn’t end up in the hands of a single company.”
He further added that the AI safety industry “is just a pretense for securing more control. If you’re interested in safety, decentralization and freedom is kind of the way to go.” It is clear that Extropic is on the side of open source and believes that closed door companies such as OpenAI, though are accelerating, are not helping in the way they sought out to do initially.
Keep following the GitHub repository of Extropic AI for updates on its models, which are coming soon.