Transformer Co-Creator Llion Jones to Address AI's Paradox and Future at TEDAISF

Llion Jones, a pivotal co-author of the seminal "Attention Is All You Need" paper and the current Chief Technology Officer and co-founder of Sakana AI, is scheduled to speak at the TEDAISF conference on October 21-22. His presentation will explore the origins of the Transformer architecture and emphasize the critical role of open-ended research in fostering significant AI breakthroughs.

The genesis of the Transformer, which underpins much of today's advanced artificial intelligence, was surprisingly organic. As announced by Sakana AI, > "Transformerのアイディアは、何気ない雑談から生まれました…世界を変えようとしたのではなく、ただ翻訳の精度をよくしたかっただけなのです" (The idea for the Transformer was born from casual conversation… we weren't trying to change the world, we just wanted to improve translation accuracy). This highlights its development from a practical need rather than a grand vision.

Jones is set to discuss the "paradoxical situation" that the Transformer's widespread success has created within the AI industry. He will advocate for open-ended research, asserting its necessity for generating major breakthroughs rather than pursuing narrowly defined objectives. His talk will also introduce new concepts and initial findings aimed at navigating and overcoming current industry challenges.

After more than a decade at Google, Jones co-founded Sakana AI in Japan with David Ha, driven by a desire to explore new AI architectures. He previously noted that "the bureaucracy had built to the point where I just felt like I couldn’t get anything done" at his former company, underscoring his move towards a more agile, research-focused environment. Sakana AI champions a philosophy of nature-inspired, adaptive AI, focusing on collective intelligence and evolutionary computing.

Sakana AI's innovative approach aims to transcend the conventional method of simply scaling existing models. Their projects, such as the "AI Scientist" which conducts autonomous research, exemplify their commitment to exploring diverse techniques. This philosophy reflects a belief that continuous learning and varied methodologies, rather than exclusive reliance on popular algorithms, will pave the way for the next generation of AI advancements.