AI startup Counterintuitive is working on a new way to build smarter machines. Their goal is to create “reasoning-native computing,” which could help AI systems understand more like humans do. Right now, most AI relies on pattern recognition, but this new approach aims to bring genuine comprehension to machines.

The Problems Holding Current AI Back

Gerard Rego, the chairman of Counterintuitive, explains that two main issues limit today’s AI systems. The first is the lack of reliable numerical foundations. Many AI models depend on outdated math methods, like floating-point arithmetic, which was designed for speed decades ago. This leads to tiny errors stacking up over time, making results unpredictable. Running the same AI twice can produce different answers, which makes verification, auditing, and reproducibility difficult. This is especially problematic in fields such as law, finance, and healthcare where accuracy is critical.

The second major problem is the architecture of current AI models. These models do not have memory. Instead of understanding or remembering why they made a decision, they just predict the next output based on patterns. It’s similar to advanced predictive text—once an output is generated, the AI can’t revisit or build on its reasoning. This limits their ability to think deeply or reason through complex problems.

Building a Team to Solve the Twin Trap

Counterintuitive is assembling a team with experts in math, physics, computer science, and engineering. Their focus is on solving these fundamental problems—the so-called “Twin Trap.” The company has more than 80 patents pending for hardware and software systems aimed at creating deterministic reasoning, causal memory, and new software frameworks. These innovations could define the next generation of AI technology.

Solving these issues could have huge benefits. If AI systems can be built with solid numerical foundations and memory, they could think and decide more like humans do. This would lead to more stable, reliable, and genuinely intelligent AI. Industries like law, finance, and healthcare could see major improvements, with AI making better decisions and providing more trustworthy insights.

A New Era for Artificial Intelligence

Counterintuitive’s approach could change how AI develops in the future. By addressing the Twin Trap at its core, the company aims to unlock a new kind of AI that’s more human-like, stable, and efficient. With a talented team and a portfolio of pending patents, they are positioning themselves to make a major impact on the field.

As AI continues to evolve, solving these fundamental problems seems essential for creating machines that can truly understand and reason. With efforts like Counterintuitive’s leading the way, the future of AI looks promising and full of potential for breakthroughs that could change many industries and how we interact with technology every day.

Inspired by

• https://www.artificialintelligence-news.com/news/ai-twin-trap-next-generation-chip-and-software/

Sources

• counterintuitive.ai
• ai-expo.net
• techexevent.com
• techforge.pub
• techforge.pub