Leading AI Expert Says Human-Level Intelligence for AI Will Take Decades
Technology experts, scholars, and industry leaders from around the world gathered in Hong Kong this week to share their insights on artificial intelligence (AI). The symposium, hosted by the Hong Kong University of Science and Technology (HKUST) and the Greater Bay Area Association of Academicians (GBAAA), featured a keynote speech by Turing Prize winner Professor Yann LeCun, considered one of the Godfathers of AI.
Contradicting concerns raised at last month’s AI Summit in Bletchley Park, where global leaders and tech specialists discussed the benefits and dangers of AI, Professor LeCun stated that it may take decades for AI to reach human-level intelligence.
The AI Summit concluded with the signing of the Bletchley Declaration, an agreement among countries including the UK, United States, and China to collectively manage the potential risks associated with AI development and deployment. However, Professor LeCun’s remarks cast doubt on the urgency expressed at the summit.
In his keynote speech, Professor LeCun also highlighted the potential uses of AI, suggesting that objective-driven AI could help predict the future and bring about a new Renaissance through the Joint Embedding Predictive Architecture (JEPA) model. Earlier this year, META announced the development of an Image Joint Embedding Predictive Architecture (I-JEPA), which has shown strong performance on various computer vision tasks.
Professor LeCun envisioned a future where AI becomes a shared infrastructure similar to the internet today. He proposed that AI platforms should be made open source, with AI assistants mediating all interactions with the digital world and serving as repositories of human knowledge and culture.
The topic of open-source AI platforms sparked a discussion among experts at the symposium in Hong Kong. While Professor LeCun and HKUST Council Chairman Professor Harry Shum saw the potential of open source to promote diversity, they acknowledged that the AI industry would be divided on the idea due to different considerations in developing Large Language Models (LLM).
The symposium, organized by HKUST and GBAAA, aimed to explore the current AI research landscape and delve into cutting-edge work. HKUST President and GBAAA Council Chair Professor Nancy Ip expressed her gratitude to the experts and scholars for sharing their valuable insights. She emphasized that AI research would lead to transformative advancements with the power to redefine how we live and work.
Amidst AI’s rapid development, concerns have been raised about the unregulated advancement of AI models. The Future of Life Institute, in a letter, expressed the growing unease and called for the development of AI to be halted until its effects and risks could be managed.
As AI continues to capture the imagination and reshape our lives, the journey towards human-level intelligence remains a challenging one. Professor LeCun’s cautious estimation of decades underscores the complexity of achieving this goal. However, the symposium in Hong Kong and the ongoing global discussions indicate a commitment to understanding and responsibly harnessing AI’s potential for the benefit of all.
In the coming years, as AI technology progresses, it is crucial for both experts and policymakers to strike a balance between innovation and regulation to ensure an AI-powered future that is safe, responsible, and beneficial to humanity.
