Two young entrepreneurs, William Chen and Guan Wang, rejected millions from Elon Musk to pursue their ambitious vision of a brain-inspired AI system that could outperform traditional AI models. Their decision has led to the creation of Sapient, a startup that has already made waves in the AI industry, surpassing some of the world’s leading AI systems in abstract reasoning tasks.
Rejecting Offers to Build a Smarter AI
When Chen and Wang, both 22, first developed their AI model, OpenChat, they didn’t expect it to attract much attention. However, to their surprise, the model went viral, catching the eye of top researchers and eventually Elon Musk. Despite the multi-million dollar offer from Musk’s company, xAI, the two decided to decline, believing that large language models (LLMs) had structural limitations. They wanted to build something far more advanced: a brain-inspired reasoning system that could better mimic human intelligence.
“We believe that AGI, or artificial general intelligence, is the future, and we want to be the ones to create it,” Chen said, referring to their ultimate goal of developing an AI system that could match or surpass human intelligence in any cognitive task.
Sapient’s AI Outperforms OpenAI and Anthropic
The breakthrough for Sapient came when their new model, built on the Hierarchical Reasoning Model (HRM) architecture, outperformed established AI systems, such as those from OpenAI and Anthropic. The HRM prototype, with just 27 million parameters, surpassed larger models in tasks that tested reasoning, problem-solving, and abstract thinking. Unlike traditional AI models, which rely on statistical predictions, HRM uses a two-part recurrent structure inspired by human cognitive processes—mixing slow, deliberate thought with fast reflexive reactions.
HRM has already demonstrated impressive capabilities in tasks like Sudoku-Extreme, navigating complex mazes, and achieving high performance on the ARC-AGI benchmark, all without the need for brute-force scaling or chain-of-thought prompting.
The Path to Artificial General Intelligence (AGI)
Chen and Wang’s ultimate ambition is to create a model capable of continuous learning, allowing the AI to absorb new experiences without retraining from scratch. They believe the future of AGI lies not in scaling up existing models but in creating smaller, more efficient architectures that can reason and plan more effectively.
The founders are optimistic that their approach will lead to AGI within the next decade. “One day, we’re going to have an AI that’s smarter than humans,” Chen said. “We want to be the ones who make it happen.”
