外滩大会今日开幕，图灵奖得主称人工智能进入“经验时代”

Core Insights - Artificial intelligence is entering an "experience era," where continuous learning will be central to its development, surpassing previous capabilities [2] - The expansion of infrastructure is facilitating the industrial scaling of AI, leading to a new "agent economy" characterized by interactions among numerous intelligent agents [3] - The rise of AI is significantly increasing global energy consumption, necessitating advancements in nuclear fusion as a sustainable energy source for future AI technologies [4] Group 1: AI Development and Learning - Richard Sutton, the Turing Award winner, emphasizes that the current machine learning methods are reaching their limits in transferring human knowledge, necessitating a new data source generated through direct interaction with the environment [2] - Sutton argues that fears surrounding AI, such as bias and job loss, are exaggerated, and that decentralized collaboration will drive human prosperity alongside AI [2] Group 2: Infrastructure and Economic Transformation - Zhang Hongjiang highlights the ongoing relevance of the "scaling law" for large models, indicating that the interaction among intelligent agents will profoundly reshape economic structures [3] - The concept of an "agent economy" is introduced, where organizations will need to enhance computational power and data richness to leverage the capabilities of intelligent agents [3] Group 3: Energy Consumption and Nuclear Fusion - Sun Xuan points out that AI currently consumes 1.5% of the Earth's electricity, with projections suggesting it could rise to over 20%, creating a significant energy gap [4] - Nuclear fusion is presented as a solution to meet the future energy demands of AI, with its high energy density being a key advantage [4] - Despite the challenges in achieving nuclear fusion, advancements in AI technology are seen as pivotal in moving towards commercial viability in this field [4]