新AI新时代：2024，AI的里程碑之年 

 王飞跃

        毫无疑问，即将过去的2024，将是智能科技与产业历史上的里程碑之年：不是一个，而是许多，许许多多代表着跨越和开创性的里程碑工作。这一年，在持续高涨的生成式语言大模型ChatGPT的热浪中开局，阳春二月又被视频大模型Sora短短的一分钟表演所震撼。紧接着，全球各种AI应用创新和产品不断迭代更替，从抖音的豆包到特斯拉的人形机器人Optimus, 一浪高于一浪，令人目不暇接。在“All in AI”和 “FOMO”的社会效应下，AI硬件巨头英伟达用了不到十个月的时间，市值就从一万亿飞涨到三万亿美元，成为全球第二大上市公司，很快又超过苹果公司，一度全球市值第一。而苹果的出路，就是推出自己的大模型，希望消费者把AI当成Apple Intelligence的缩写，利用生成式智能，继续维持其产品的领先地位。

        年终的结局更是壮观，不但物理和化学的诺贝尔桂冠都落到非物理非化学专业的人工智能研究者或创业者头上，就连一幅机器人创作的人工智能之父图灵的头像，居然也让苏富比以“AI God”之名，拍出百万美元的天价，高于预期近十倍。特别是“全知”型第二版的ChatGPTo3的发布会，从12月4日起，用了12个工作日连播到圣诞节，向全世界展示其功能已比三个月前的第一版有了惊人的提高，不但证明了Scaling Law还有巨大潜力，更为大模型的实用化带来了无限的想象。

        例如，o3版的智商号称157，与爱因斯坦的估值不相上下，高于99%的人类。这是由o3在国际编程基准CodeForces的排名而推算出来的，其在数学基准FrontierMath，抽象推理基准ARC-AGI，博士水平科学基准（CPQADiamond）等开源数据库的测试结果也远超人类平均水平。虽然这与AGI真正要求显然仍有差距，但进步显著，已让大厂的程序员担心失业，甚至让AI顶会NeurIPS上的研究生后悔读博，更为明年AI更高的浪潮和更高的期望打下基础。

        这一切预示着什么？你可能会All in AI！可是AI又是什么？

        首先，一场人类社会知识的基础设施大变革已经开始，专业知识的定义与内涵即将改变，并将冲击社会的行行业业各个角落。今日之专业知识，明日将成为科学素质和科技常识，不学也要会，不行就去问！而且，大部分内容不再需要课堂式的传授。据报道，美国亚利桑那州今年己推出完全由AI教授学术课程的学校。海外网上关于“University is dead”,甚至“Education is dead”的言论，虽然偏激，但有必须思考的道理。其次，AI的含义也变了，Artificial Intelligence已成为老AI“Old AI"。从2024年起，大模型已将AI重铸为“Agentic Intelligence"，但这也将很快成为旧AI“Past AI”。美国云软件巨头Salesforce声称:“要么AI，要么灭亡”，其AI就是“Past AI”，尽管它的核心产品Agentforce从1.0到2.0的升级，用了不到3个月的时间。智能科技发展的趋势清楚地表明，时代的AI不久就会嬗变为“Autonomous Intelligence”，成为集无人机、无人车、无人船等各种物理和数字机器人为一体的新AI“New AI”。显然，我国正在兴起的低空经济及生态产业，将是“新AI”重大的应用场景。

        届时，“老、旧、新”AI将协同发挥作用，构成混合式平行智能Parallel Intelligence，大步迈向知识自动化Knowledge Automation产业，揭示AI之本质，就是Dartmouth第一次人工智能会议的发起人John McCarthy教授心中的AI：Automation of Intelligence。这将使另一位人工智能之父，诺贝尔奖获得者Herbert Simon教授认为的“原本无法传递、难以制造、不可能成为社会商品”的“Trust”与“Attention”，实实在在地由新AI技术转化为智业社会里可大规模生产、大批量流通的基础性新商品，不但创新了商品的范畴，扩大了经济空间，同时拓广了提效途径，确保了智能经济的边际效用递增之效果，让世界走上新型可持续发展的智慧之道。

        新AI对于就业的冲击，是我们必须关注的重大社会问题。短期而言，应规范推动PAPAI++新行业的发展，即提示工程师（Prompt Engineer）、对齐工程师（Alignment Engineer）、引导工程师（Prescription Engineer）等面向“老、旧、新”AI产业的新工种。长期而言，新AI必将升级人类的就业方式，使大量昔日非工作的活动成为新的工作方式，不但不会让人失业，还会比机器计算机等更加有效的为我们提供更多更好更可持续的新工作岗位。当下的“快递小哥”、“网红卖货”, 只是未来新工钟的“原始形态”而已，却已经表明新AI不会更不可能让人类失业。而且，新技术的发展是历史性潮流，浩浩荡荡，不是一个人甚至一个国家可以阻挡的。我们必须积极应对，主动有效地研究治理变革过程中衍生出来的各种相关问题。

        正如联合国秘书长古特雷斯近日在联合国安理会关于AI与维护国际和平和安全的高级别会议上所言：无论如何，“人类命运不能交由算法决定”，不管算法如何智能，是“老AI”还是“新AI”。从理论和技术上，对于特定问题与特别领域，可解释的AI是必然的追求；但对于通用智能，相对于人类的认知能力，智能本质上是不可解释的，无论在其前面加多少形容词也不行。因此，AI和智能科技的治理变得极其重要，必须人类主导，这将催生新的社会形态与生态，世界上一些主要国家已认识到这个问题。欧洲更是跑在前列，今年八月实施的欧盟AI法案，以及之前启动的数据等相关法案，如GDPR，值得我们关注与研究。

        纵观全球AI的现状与发展趋势，三个方向将从2025起加速展开：首先是数字人智能技术，其中AI for Science (AI4S)将在AI的诺贝尔桂冠加持下向各个学科深入渗透，有望实现从AI Agents到数字人、机器人、生物人一体化，具有平行智能Parallel Intelligence的平行科学家Parallel Scientists之跨跃性突破，并冲击各行各业，得到广泛的应用；其次是组织化可运用的自主智能团体(Autonomous Smart Organizations, ASO)技术的突破，使每个人都可以具有自己的数字人公司或组织的愿景成为现实，极大的扩展了个人的能力与执行力，同时加剧社会的二极分化现象，社会对个人的教育水平要求会越来越高，读博无须后悔，否则很快又可能再次真正的后悔；最后，借助区块链智能，智能合约，DAO（Decentralized Autonomous Organizations with Decentralized Autonomous Operations）等新兴技术，AI治理将走上硬核技术化并广泛应用，无论是今年中东发生的Pager Bombs事件（证明了未来的用品，从原材料到中间构件，必须利用区块链和智能合约等技术进行加密、追踪、使用，否则将成为社会的重大安全隐患），还是当下海外风行的DigitalNomads运动（年轻人利用加密币和DAO进行生活与创业，不读我们的书，不打我们的工，不用我们的钱，不交我们的税，高效率低成本，不到十年其经济规模已愈五万多亿美元，己差不多占全球GDP的5%之巨），都在警示我们必须突破传统的治理思维，治理也是产业，极其重要的产业，必须有效益，治理已不再是传统文科的领地，更需要AI和智能技术。

        在此形势下，中国的AI科技及产业大有作为，新质生产力更是有了更大的作用天地。但我们必须正视目前与国际前沿AI技术与产业的差距，首先变革自己的思维方式，然后实事求是，在迎头赶进的同时，加大基础科研的支持力度。实际上，对于AI基础研究而言，由于算力和Scaling Law的影响，除了极少机构，大多数科研究人员己难以获得一线参与实战的机会，就像百年前研究物理却没有原子实验设备和量子物理理论知识一般，这己成学界的全球性问题。这一现象是短暂的，但十分关键，特别是对中国这样的大国，必须认真应对。各种AI4S方法, 特别是DeSCI4S  (Decentralized Science, Computation, and Intelligence for Sciences), AI4AS (Autonomous Intelligence for Autonomous Systems), SI4SI (Systems Intelligence for Sustainability Industry), 包括各种计算实验的无人工厂化，有效利用并充分发挥可持续性能源的优势，切实打通“小知识、大数据、深智能”之循环因果(Circular Causality)的数据化产业化道路，值得我们深入思考，加快行动。

        目前的国际形势和产业现状表明，AI技术突破的机会依然很多，技术的迭代更是难以垄断，无论算法还是算力，中国一定能够达到并取得领先世界的水平，特别是我们有巨大的科技人力和绿色能源优势，但前提是尊重规律，不可投机，长远规划，另辟新径，直道行车，做好自己的事，坚持全球化合作，坚持造福全人类。

 An Era of Broad Societal Shift 

From the meteoric rise of generative language models such as ChatGPT to the shockwave caused by video model Sora's performance in February 2024, to the rapidly expanding market share of NVIDIA, the year 2024 witnessed an array of breakthrough developments in artificial intelligence technology and its applications, signaling a transformation that is reshaping industries, economies and societies. 

Apple's proprietary AI is positioning the tech giant to continue its leadership in consumer products, while others, such as Tesla's humanoid robot Optimus and TikTok's Doubao — continue to push the boundaries of AI in the physical world.

But perhaps the most profound aspect of the year was the emergence of AI as an ever-present force. The Nobel Prizes in Physics and Chemistry, typically awarded for groundbreaking discoveries in traditional scientific fields, were bestowed upon AI researchers and entrepreneurs in 2024. More interestingly, a portrait of AI pioneer Alan Turing created by a robot sold at Sotheby's for nearly 10 times its expected value, showing how deeply AI has permeated the cultural and intellectual zeitgeist.

This rapid advancement is beginning to unsettle professionals in the tech industry.

The definition of "professional knowledge" is undergoing a radical shift. What was once exclusive expertise in fields such as law, medicine and engineering will soon become a set of essential skills accessible to everyone. As reports indicate, a new school in the US state of Arizona is now teaching courses entirely via AI.

This transformation signifies the arrival of a new era — one in which AI is rapidly becoming a core element of everyday life. As generative models evolve, we are transitioning from what we once called "artificial intelligence" to "agentic AI".

The reality is, AI's reach is expanding across all domains, and the implications are profound. From autonomous vehicles to unmanned aerial drones, the future of "New AI "is already taking shape. China's burgeoning low-altitude economy, along with its ecological industries, represents one of the most significant application areas for this new wave of AI.

At the heart of this transformation lies a fundamental question: What does AI mean for the future of work? As AI continues to evolve, it will not simply replace existing jobs — it will redefine them. In the short term, emerging job roles such as prompt engineers, alignment engineers, and prescription engineers are taking shape to support the growing "old", "past", and "new" AI industries. These roles will enable workers to interface with AI technologies and ensure they align with human values.

In the long term, however, AI will bring about a more profound change: it will redefine the very concept of employment, with tasks that once seemed mundane or outside the realm of traditional work becoming essential components of a new, AI-driven economy.

Furthermore, the rise of AI and intelligent technologies signals a broader societal shift — one that extends beyond the economy and touches on governance, security and international relations. As AI becomes increasingly integrated into daily life, we must also address the governance challenges it presents. AI models must be understood and controlled, and this requires a strong regulatory framework that prioritizes transparency, ethics and human oversight.

As we consider these transformations, it is important to recognize that the governance of AI is not just a technical issue; it is a matter of global importance. As UN Secretary-General António Guterres recently remarked humanity's fate can't be left to algorithms.

To ensure that AI serves the common good, we must establish global norms and regulations that keep humanity at the center of the AI conversation. Economies around the world are beginning to take notice. The European Union AI Act, which came into force in August 2024, sets a precedent for how AI technologies should be regulated to safeguard public interests.

Looking ahead, three key areas will accelerate from 2025. First, AI for Science (AI4S) and digital human intelligence will penetrate all disciplines, creating "parallel scientists" by integrating AI agents, digital humans, robots and bio-humans, transforming industries.

Second, breakthroughs in autonomous smart organizations will allow individuals to create digital-human-powered companies, greatly enhancing personal capabilities. This will increase societal polarization, with a growing demand for higher education.

And third, AI governance will increasingly rely on blockchain, smart contracts and decentralized autonomous organizations, shifting governance into the realm of hard technology. Events such as the Middle East's "pager bombs" and the rise of the "digital nomads" movement highlight the need to rethink governance and industry practices. AI and intelligent technology will drive this transformation.

China's AI industry holds immense potential to make a significant global impact. However, there is a gap between China and global AI leaders. To bridge this gap, we must shift our mindset, focus on pragmatic progress and accelerate basic research while staying grounded. This requires respecting development laws, avoiding opportunism, planning for the long term, adopting new approaches, and focusing on global cooperation for the benefit of all.