Anna Goldie and Azalia Mirhoseini, the brains behind Ricursive Intelligence, share a remarkable journey in the AI world. Both known figures in the industry, they turned down tempting offers from tech giants like Zuckerberg. Their story began at Google Brain, where they created the Alpha Chip, a tool that redesigned chip layouts in hours instead of the typical year-long task.
Just four months after founding Ricursive, they secured a whopping $300 million in Series A funding, raising their company’s valuation to $4 billion. Their unique approach? They design AI tools for chip creation rather than competing as another chip manufacturer like Nvidia. In fact, Nvidia is among their investors, along with other major chipmakers who could benefit from their technology.
Goldie explains, “We want to enable any kind of chip to be built quickly and efficiently using AI.” Their collaboration goes way back to their days at Stanford, where Goldie was a PhD candidate and Mirhoseini taught computer science. Their teamwork has been synchronous, from Google Brain to Anthropic, and now at Ricursive.
Working closely at Google, their bond grew even stronger; they even shared workout sessions. This camaraderie led to playful nicknames, reflecting both their professional and personal connection. Yet, their rise hasn’t been without challenges. Their groundbreaking work attracted critiques, including a colleague who publicly discredited their contributions, despite their significant role in Google’s chip design.
Designing chips is complex. Engineers must carefully place millions of tiny components on silicon. But with the Alpha Chip, complex designs could be drafted in about six hours. Goldie emphasizes how this tool learns with every iteration, improving efficiency with experience.
Ricursive aims to take things further. Their platform can learn from multiple designs, enhancing future chip creations. By integrating large language models, they plan to streamline the entire design process. This could accelerate advancements in AI, which require ever-improving hardware.
Goldie believes that better chip design is crucial for AI’s evolution. “Chips are the fuel for AI,” she says. Mirhoseini highlights how the lengthy traditional design process slows down AI progress. Their goal is to enable rapid improvements in hardware, which could lead to increased AI efficiency without exhausting global resources.
Imagine AI able to design its own optimized brains, improving performance and reducing costs dramatically. “We could achieve almost a 10x improvement in performance per total cost of ownership,” Goldie points out. Their early conversations with major chip manufacturers indicate a strong interest in their innovative approach.
As Ricursive develops, they aim to push the boundaries of what’s possible with AI and chip design. The future looks bright, and their vision could reshape the tech landscape. For those interested in further insights into the world of AI chip design, the importance of efficient hardware, and the potential for future advancements, check out Wired’s coverage for more context.
Source link
AI chips,Ricursive Intelligence
