全球著名华人人工智能学者李飞飞联合领导的斯坦福大学以人为本人工智能研究所（Stanford HAI）发布了《2024 年人工智能指数报告》（Artificial Intelligence Index Report 2024）。Stanford HAI 官方介绍道，“这是我们迄今为止最全面的报告，而且是在人工智能对社会的影响从未如此明显的重要时刻发布的。”
2023人工智能发展10大趋势
1
人工智能在某些任务上胜过人类，
但并非在所有任务上
AI beats humans on some tasks, but not on all.
人工智能已在多项基准测试中超越人类，包括图像分类、视觉推理和英语理解等方面。然而，在竞赛级数学、视觉常识推理和规划等更复杂的任务上，人工智能却落后于人类。
AI has surpassed human performance on several benchmarks, including some in image classification, visual reasoning, and English understanding. Yet it trails behind on more complex tasks like competition-level mathematics, visual commonsense reasoning and planning.
2
产业界继续主导人工智能前沿研究
Industry continues to dominate frontier AI research.
2023 年，产业界产生了51个著名的机器学习模型，而学术界仅贡献了 15 个。2023 年，产学合作还产生了 21 个著名模型，创下新高。
In 2023, industry produced 51 notablemachine learning models, while academia contributed only 15. There were also 21 notable models resulting fromindustry-academia collaborations in 2023, a new high.
3
前沿模型变得更加昂贵
Frontier models get way more expensive.
根据 AI Index 的估计，最先进人工智能模型的训练成本已达到前所未有的水平。例如，OpenAI的GPT-4估计使用了价值7800万美元的算力资源进行训练，而谷歌的Gemini Ultra则花费了1.91亿美元的算力资源。
According to AI Index estimates, the training costsof state-of-the-art AI models have reached unprecedented levels. For example, OpenAI’s GPT-4 used anestimated $78 million worth of compute to train, while Google’s Gemini Ultra cost $191 million for compute.
4
美国领先于中国、欧盟和英国，
成为顶级人工智能模型的主要来源
The United States leads China, the EU, and the U.K. as the
leading source of top AImodels.
2023 年，61 个著名的人工智能模型源自美国的机构，远远超过欧盟的 21 个和中国的 15 个。
In 2023, 61 notable AI models originated from U.S.-based institutions, far outpacing the EuropeanUnion’s 21 and China’s 15.
5
严重缺乏对 LLM 责任的可靠和
标准化评估
Robust and standardized evaluations for LLM
responsibility are seriously lacking.
包括 OpenAI、谷歌和 Anthropic 在内的领先开发商主要根据不同的负责任人工智能基准测试其模型。这种做法使得系统地比较顶级人工智能模型的风险和局限性的工作变得更加复杂。
New research from the AI Index reveals a significant lack of standardization in responsible AI reporting.Leading developers, including OpenAI, Google, and Anthropic, primarily test their models against differentresponsible AI benchmarks. This practice complicates efforts to systematically compare the risks andlimitations of top AI models.
6
生成式人工智能投资激增
Generative AI investment skyrockets.
尽管去年整体人工智能私人投资有所下降，但用于生成式人工智能的资金激增，比2022年增长了近八倍，达到252亿美元。生成式人工智能领域的主要参与者，包括OpenAI、Anthropic、Hugging Face和Inflection，都报告了可观的融资轮次。
Despite a decline in overall AI private investment lastyear, funding for generative AI surged, nearly octupling from 2022 to reach $25.2 billion. Major players inthe generative AI space, including OpenAI, Anthropic, Hugging Face, and Inflection, reported substantialfundraising rounds.
7
数据显示，人工智能让打工人更有生产力，
工作质量更高。
The data is in: AI makes workers more productive and
leads to higher quality work.
2023 年，多项研究评估了人工智能对劳动力的影响，表明人工智能使工人能够更快地完成任务，并提高产出质量。这些研究还表明，人工智能有可能缩小低技能和高技能工人之间的技能差距。不过，其他研究也提醒说，在没有适当监督的情况下使用人工智能可能会导致绩效下降。
In2023, several studies assessed AI’s impact on labor, suggesting that AI enables workers to complete tasks morequickly and to improve the quality of their output. These studies also demonstrated AI’s potential to bridgethe skill gap between low- and high-skilled workers. Still, other studies caution that using AI without properoversight can lead to diminished performance.
8
得益于人工智能，科学进步进一步加速
Scientific progress accelerates even further, thanks to AI.
2022 年，人工智能开始推动科学发现。2023年，与科学相关的更重要的人工智能应用启动--从使算法排序更高效的AlphaDev，到促进材料发现过程的GNoME。
In 2022, AI began to advancescientific discovery. 2023, however, saw the launch of even more significant science-related AI applications—from AlphaDev, which makes algorithmic sorting more efficient, to GNoME, which facilitates the process ofmaterials discovery.
9
美国的人工智能法规数量急剧增加
The number of AI regulations in the United States sharply increases.
美国人工智能相关法规的数量在过去一年和过去五年中大幅增加。2023 年，人工智能相关法规将从 2016 年的 1 项增加到 25 项。仅去年一年，人工智能相关法规的总数就增长了 56.3%。
The number of AIrelatedregulations in the U.S. has risen significantly in the past year and over the last five years. In 2023, therewere 25 AI-related regulations, up from just one in 2016. Last year alone, the total number of AI-related regulationsgrew by 56.3%.
