TECHNOLOGY · ANALYSIS
Ricursive aims to put AI in charge of designing entire chips
The founders of Ricursive Intelligence present an approach to AI chip design that learns from successive iterations. They say they optimize chips as a whole. Their appearance outlines an industrial strategy without establishing measured gains for the company’s future products.
Legal disclosure: this content was produced by an artificial intelligence system. Editorial transparency
Ricursive Intelligence is developing AI intended to design chips, learn from each iteration and improve subsequent designs. Anna Goldie and Azalia Mirhoseini presented this approach during their appearance at TechCrunch Disrupt on Wednesday, according to TechCrunch’s account. Their stated goal is to automate and accelerate the design process.
The founders say they are working on optimizing entire chips, rather than dividing them into smaller blocks under the traditional approach described in the article. They explain that this method allows AI to produce complex layouts, including wavy patterns. These capabilities remain claims by the executives: the account provides no quantified comparison of the performance of Ricursive’s chips.
They also cite AlphaChip, a system they had previously developed. According to their account, it had proposed a circular arrangement of components that initially surprised engineers and was later judged successful. This earlier example is insufficient to demonstrate their new company’s results.
The potential industrial benefit of the approach would be to make more specialized chips economically feasible by reducing the design effort. That prospect depends, however, on the gains actually achieved and their translation into manufactured components. The information presented therefore establishes a technical direction and an ambition, without demonstrating that they will resolve computing capacity constraints.