Forget Nvidia: Ndea wants to build AI that keeps improving on its own with ‘no bottlenecks in sight’

François Chollet, a former Google engineer and the creator of the widely used Python deep learning framework Keras, has co-founded Ndea, a new AI research and science lab, alongside Mike Knoop, co-founder of Zapier.

In a post on the startup’s new website, the founders explain their goals of combining intuitive pattern recognition, enabled by deep learning, with formal reasoning through what they call “guided program synthesis.”

According to them, this fusion will allow AI systems to adapt and innovate far beyond current task-specific applications, ultimately leading to artificial general intelligence (AGI), defined loosely throughout the AI community as machine intelligence that can outperform human beings at most economically valuable and cognitive tasks. As they write:

“We need computers that can pose problems and explore new territory, not just apply known solutions. We need computers that can innovate. The path to AGI is not through incremental improvements to existing methods.”

The duo hasn’t stated yet whether or not they’ve received outside funding for this venture or are bootstrapping it with their own funds.

It comes several months of after former OpenAI co-founder and chief scientist Ilya Sutskever, reported to have led the briefly successful yet ultimately reversed internal coup against his fellow co-founder, Sam Altman, also announced a startup focused on developing “Safe Superintelligence” with $1 billion in private backing.

Beyond deep learning

While existing deep learning systems are impressive, Chollet and Knoop argue they are fundamentally constrained by their reliance on large datasets and their inability to adapt efficiently to new tasks.

Chollet and Knoop believe that program synthesis is the key to overcoming these limitations.

Unlike traditional deep learning, which interpolates between data points, program synthesis searches for discrete programs that explain data. This method allows for much greater generalization with far fewer data points.

Combining deep learning’s intuitive capabilities with the rigorous reasoning of program synthesis could lead to a new paradigm for AI research.

“Ndea’s mission is to operationalize AGI to realize unprecedented scientific progress for the benefit of all current and future generations,” they note.

Building a “Factory for Scientific Advancement”

Ndea’s long-term vision goes beyond the creation of AGI. The lab aims to act as a “factory for rapid scientific advancement,” capable of solving both known and unknown challenges.

From tackling current frontiers like autonomous vehicles and sustainable energy to accelerating entirely new discoveries, the lab sees itself as a catalyst for scientific progress.

Chollet added that their research direction has the potential to unlock breakthroughs and redefine the boundaries of human knowledge. As he wrote in a thread on X: “If we’re successful, we won’t stop at AI. With this technology in hand, we want to tackle every scientific problem it can solve. We see accelerating scientific progress as the most exciting application of  AI.”

According to Chollet, this progress hinges on developing AI that can learn as efficiently as humans and continue to improve over time without bottlenecks.

While acknowledging that success is not guaranteed, Chollet emphasized the importance of pursuing this ambitious goal, stating on X: “We believe we have a small but real chance of achieving a breakthrough—creating AI that can keep improving over time with no bottlenecks in sight.”

A New Research Focus for AGI

Program synthesis, the cornerstone of Ndea’s research, is still a relatively young field. Chollet likened its current state to where deep learning was in 2012.

However, he noted that its potential is being increasingly recognized by frontier AI labs, even if most see it as only a small component of what’s needed for AGI.

Ndea, in contrast, considers program synthesis equally as important as deep learning and has made it central to their approach.

The lab is also actively recruiting a globally distributed team of researchers and engineers to build what it describes as the most “talent-dense program synthesis team” in the world.

The company operates as a fully remote organization and is looking for candidates with strong technical expertise, particularly in translating mathematical concepts into code.

Founders with strong track records

François Chollet and Mike Knoop bring extensive experience to Ndea.

At Google, Chollet worked on core research into deep learning and AI systems, gaining insights into the limitations of existing models and opportunities for improvement. His contributions include not only Keras but also the ARC-AGI benchmark, a widely used metric for measuring progress toward AGI.

He is also the author of the book Deep Learning with Python and has been recognized among Time’s “100 Most Influential People in AI.”

Knoop co-founded Zapier, the world’s largest AI automation company, where he led engineering and product development as well as the company’s early adoption of AI technologies.

He is also credited with pioneering best practices for globally distributed teams. Both Chollet and Knoop are co-founders of the ARC Prize Foundation, a nonprofit organization focused on advancing open AGI research.

Future visions rooted in ancient tradition

Ndea derives its name from the Greek concepts ennoia (intuitive understanding) and dianoia (logical reasoning), reflecting the lab’s goal of merging deep learning and program synthesis. By operationalizing AGI, Ndea hopes to compress centuries of scientific progress into decades or even years.

While acknowledging the uncertainty and challenges of pursuing AGI, Chollet and Knoop remain optimistic about their approach. They see AGI as the gateway to addressing humanity’s most pressing challenges and uncovering entirely new opportunities for discovery.