← Retour au blog
tech 23 May 2026

Rethinking AI: A New Architectural Approach

With massive investments in AI, is the industry betting on the right horse? Discover an innovative architectural alternative.

Article inspired by the original source
I’m writing again ↗ www.cringely.com

AI on the Rise: A Necessary Context

Artificial Intelligence (AI) continues to dominate global tech and economic conversations. According to a McKinsey report, global AI investments are expected to reach $500 billion by 2024. This digital gold rush is primarily driven by promises of increased efficiency, cost reductions, and groundbreaking innovations.

However, despite this optimism, some critical voices are emerging. Robert X. Cringely, a tech veteran, recently resurfaced with a bold perspective: the current architecture of AI systems might be misguided.

An Architectural Alternative: Why Now?

Cringely, with his company 2Brains, has filed a patent for a new architectural approach. This innovation aims to address perceived weaknesses in current systems. Today, most AI relies on massive neural network models, which are data and computation-intensive.

Cringely's proposed architecture promises to improve efficiency and reduce reliance on massive datasets. Indeed, according to a Gartner study, 80% of enterprise data is unstructured and hard to exploit by current systems, limiting their utility.

Challenges of Current Architectures

Traditional AI systems face several hurdles:

  1. Energy Consumption: A University of Massachusetts study revealed that training an AI model can produce up to 284 tons of CO2, equivalent to five gasoline cars over their lifetimes.
  1. Scale Dependency: Large companies invest heavily in cloud infrastructure to support these models, which is not viable for all businesses.
  1. Data Quality: Models require clean and structured data, often a scarce resource.

The 2Brains Approach: A Viable Solution?

2Brains' proposal stands out with a modular and adaptable approach. By using micro-models, their architecture could potentially reduce energy consumption and increase precision on smaller, specialized datasets.

This strategy could transform AI into a more accessible solution for SMEs, often excluded due to the prohibitive costs associated with current solutions.

Use Cases and Potential Impact

Imagine a logistics company aiming to optimize delivery routes. With the traditional approach, it might require millions of traffic data points to train an effective model. 2Brains' proposal could potentially achieve similar results with a subset of data, reducing resource consumption and deployment time.

Conclusion: The Future of AI

The architecture proposed by Cringely and his team might mark a turning point in AI evolution. As companies and developers become aware of the limitations of current models, adopting more efficient and sustainable solutions will be crucial.

For those looking to explore these new possibilities, we offer to discuss your project in 15 minutes. Discover how this innovation could transform your business.

Let's discuss your project in 15 minutes.

IA innovation technologique architecture IA efficacité énergétique développement durable
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call