TL;DR:
- Llion Jones, co-author of an influential 2017 paper, leaves Google to co-found Sakana AI with David Ha.
- Sakana AI, based in Tokyo, focuses on biomimicry to create adaptable and cost-effective AI models.
- Approach involves the collaboration of smaller AI models, inspired by collective intelligence in nature.
- Sakana AI aims to outperform larger systems in terms of efficiency and flexibility.
- Tokyo’s technical infrastructure positions it as a hub for Sakana AI’s innovative growth.
- Startup challenges industry giants’ rigid AI models, proposes evolutionary computing principles.
- Jones and Ha’s departure signifies a shift towards agile innovation outside corporate structures.
Main AI News:
The landscape of generative artificial intelligence (gen AI) is undergoing a rapid and dynamic evolution, marked not only by direct competition among tech giants like OpenAI, Meta, and Google in the realm of foundation models, but also by a trend where some of the most brilliant minds within these organizations are departing to establish their own gen AI startups.
In a significant departure last month, Llion Jones, a co-author of the groundbreaking 2017 research paper titled “Attention Is All You Need,” laid the foundation for the generative AI revolution by introducing the transformative architecture of transformers – a core component in leading large language models (LLMs), announced his departure from Google to embark on a new entrepreneurial journey.
This week, the curtain has been lifted to reveal his latest endeavor: Sakana AI, a pioneering AI venture headquartered in Tokyo. Collaborating as a co-founder is David Ha, the former head of research at Stability AI and a distinguished ex-researcher of Google Brain.
At Sakana, their vision takes an audacious turn, as they embrace a biomimicry approach that draws inspiration from the collective intelligence observed in natural systems such as schools of fish and beehives. This philosophy aims to engineer AI models that exhibit flexibility, reactivity, and economic efficiency – attributes drawn from the intricate dance of nature’s own systems.
Jones envisions that Sakana AI will leverage this innovative strategy to construct an avant-garde generative AI model adept at producing diverse content forms, ranging from text and images to code and multimedia compositions. While the startup has remained tight-lipped about its backers, opting not to disclose this information to Reuters, the anticipation for Sakana AI’s trajectory is palpable.
Revolutionizing AI with a Natural Touch
Sakana AI stands apart with a revolutionary strategy that involves the orchestration of multiple smaller AI models to work in tandem, analogous to the harmony seen in a swarm, yielding intricate outcomes. This methodology counters the conventional trend of constructing monolithic AI systems.
Propelled by its swarm-based paradigm, Sakana AI is confident that its approach, inspired by the principles of collective intelligence, can achieve outcomes on par with larger systems, all while offering enhanced cost-effectiveness and adaptability.
Speaking with Bloomberg, David Ha likened this approach to the remarkable way ants dynamically form bridges, showcasing rapid adaptation in the face of change. This resilience, Ha believes, holds significant promise as a guiding principle for building algorithms inspired by nature.
The Birth of “Sakana”
The name “Sakana” traces its roots to the Japanese word for fish (さかな). It encapsulates the idea of diverse entities converging to form a coherent whole through simple yet elegant rules. As explained to The Financial Times by the founders themselves, the term captures the essence of nature’s intricate patterns reflected in their approach.
Tokyo: A Thriving AI Haven
Sakana AI’s choice to establish its headquarters in Tokyo is strategic, with the city boasting advanced technical infrastructure and a highly skilled pool of talent. This positioning aligns with the startup’s mission to harness Tokyo’s potential in nurturing AI innovation, attracting global expertise, and fostering research and development. While the specifics of their office space remain undisclosed, the strategic move speaks volumes about their dedication to innovation and collaboration.
Redefining AI’s Trajectory
Sakana AI’s vision extends beyond the creation of mere AI models. Co-founders Ha and Jones are adamant about addressing the shortcomings inherent in many contemporary AI systems, which often suffer from rigidity and inflexibility. Their proposition revolves around AI models that embody evolutionary computing principles, drawing inspiration from nature’s remarkable adaptability.
Jones and Ha challenge Google’s singular focus on large language models, expressing their belief that this myopic approach is limiting. While they remain open to utilizing such models, they see the potential for a more diverse and adaptable AI landscape.
The Shift to a New Era
The transition of Jones and Ha from tech titans to the helm of Sakana AI marks a significant milestone in the realm of generative AI. The move represents a divergence from the constraints of corporate bureaucracy and a quest for more impactful innovation. Just as they challenge the status quo, Ha also raises concerns about the current state of AI giants like OpenAI, suggesting that a lack of widespread development-sharing may impede true innovation.
Conclusion:
The emergence of Sakana AI, driven by the co-authors of a seminal AI research paper, marks a turning point in the generative AI landscape. By harnessing nature’s wisdom and orchestrating collaborative AI models, Sakana AI not only seeks efficiency and adaptability but also challenges the conventional paradigms set by industry titans. As Tokyo provides a conducive environment for innovation, Sakana AI’s approach could influence the market’s trajectory towards more flexible and nature-inspired AI solutions.
