(Note: This article was motivated by Tech Crunch’s piece written by Rebecca Bellan.)
Mira Murati, has named her new company “Thinking Machines Lab” with a valuation of $10B dollars and on a seed round of $2B. Led by Andreessen Horowitz, with participation from Sarah Guo’s Conviction Partners, the raise doubled the company’s original $1 billion goal. According to Business Insider, investors had to commit at least $50 million to participate.
Mira Murati’s track record and talent pool she can choose from makes Thinking Machines and expected success. Another reason for the expected success is the focus on the fields of science and engineering, but in building these solutions they are looking to make the AI system “widely understood, customizable, and generally capable,” and close major gaps in in today’s models and prioritization.
Mira Murati’s credibility is beyond reproach having led the development of at OpenAI on ChatGPT, DALL E and GitHubs’s Codex (called Copilot). Besides herself the founding team consists of rock stars in their own right. John Schulman, an OpenAI co-founder and co-creator of ChatGPT, is chief scientist. Barret Zoph, former VP of research at OpenAI, is CTO. Other key hires include Jonathan Lachman, Lilian Weng, Luke Metz, Sam Shleifer, and Stephen Roller—many of them with OpenAI ties. Researchers from DeepMind, Meta, CharacterAI, and Mistral AI have also joined. Bob McGrew and Alec Radford, both key figures in OpenAI’s early breakthroughs, are advisors.
A key question in an evaluation this high is, what makes Thinking Machines unique? I mean, we have enough people improving search with AI, but being me too doesn’t mean much, and given Murati’s previous interviews, you would expect a great deal of strategy. On ThinkingMachines.ai we have something that looks like a manifesto typed on an old Smith-Corona typewriter.
The stated purpose is “building a future where everyone has access to the knowledge and tools to make AI work for their unique needs and goals.”
After supporting paragraphs come a series of goals or best practices
- Science is better when shared: “We think sharing our work will not only benefit the public but also improve our own research culture.
- AI that works for everyone: “We are excited to build multimodal systems that work with people collaboratively … and can adapt to the full spectrum of human expertise and enable a broader spectrum of applications.”
- Solid foundations matter: “Model intelligence as the cornerstone. Ultimately, the most advanced models will unlock the most transformative applications and benefits, such as enabling novel scientific discoveries and engineering breakthroughs.”
“Infrastructure quality as a top priority. Research productivity is paramount and heavily depends on the reliability, efficiency, and ease of use of infrastructure. We aim to build things correctly for the long haul, to maximize both productivity and security, rather than taking shortcuts.
- Learning by doing: Products enable iterative learning through deployment, while great products and research strengthen each other. Products keep us grounded in reality and guide us to solve the most impactful problems.”
“The most effective safety measures come from a combination of proactive research and careful real-world testing. We plan to contribute to AI safety by
- (1) maintaining a high safety bar--preventing misuse of our released models while maximizing users' freedom,
- (2) sharing best practices and recipes for how to build safe AI systems with the industry, and
- (3) accelerating external research on alignment by sharing code, datasets, and model specs.
“We believe that methods developed for present day systems such as effective red-teaming and post-deployment monitoring, provide valuable insights that will extend to future, more capable systems.”
- Join Us: You can follow us on X at @thinkymachines, or submit job applications here if you're interested in working with us.
- Product: Join us in the exciting early stages of building something transformative. We are looking for people with a strong track record of building successful AI-driven products from the ground up and enthusiasm about wearing multiple hats.
- Core Infrastructure: We're looking for engineers with deep experience in building and maintaining fault-tolerance, scalability, and secure infrastructure via load balancing, autoscaling, monitoring, and various elements for service orchestration.
- Machine Learning: Whether you hold a PhD or are self-taught, we're interested in candidates who can demonstrate concrete achievements in ML research and engineering through.
Reading the manifesto in full I am not sure if I am going to a revival meeting or an episode of HBO’s Silicon Valley.
However, “Building next-gen models isn’t cheap. Training large systems and hiring top talent requires serious capital. In that sense, the $2 billion isn’t about runway—it’s table stakes.”
Edited by
Erik Linask
