阿波罗全球管理公司首席经济学家托尔斯滕·斯洛克近日发布了一份措辞尖锐的报告，直言所谓“人工智能革命”的真实波及范围远不如宣传中那么广泛。数据显示，人工智能对宏观经济的影响目前依然十分有限，这可能意味着在金融趋势和全球经济走向方面，我们将迎来一段相当耐人寻味的时期。
根据斯洛克在《Daily Spark》专栏中的分析，尽管围绕大语言模型研究的最新进展投入了巨额资金、也伴随着巨大声浪，但到目前为止，人工智能在美国科技行业之外的企业界并未带来任何财务回报。少数与科技相关的公司确实获得了可观的收益，但标普500指数中的绝大多数企业，在其生成式人工智能或大语言模型的投入上几乎看不到任何成果。
斯洛克指出，经通胀调整后的人均营收只有“七大科技巨头”在上升，这一群体是标普500指数中科技权重最高的部分，包括英伟达、特斯拉、亚马逊、微软、Google母公司Alphabet、苹果和Meta。这一群体的利润率同样在走高。而他把指数中其余公司称为“标普493”，这些公司的人均营收和利润率则双双持平。
此外，这位阿波罗的经济学家没有看到任何人工智能推动生产率增长的迹象。简单来说，生成式人工智能和大语言模型对科技行业以外的企业利润毫无影响。由于拿不出任何实际成果，主流标普500企业已经开始收缩人工智能所需的巨额开支，把词元消耗视为拖累利润率的成本，而不再把它当作未来增长的潜在来源。
风险投资支持的公司不断加大在人工智能词元上的支出，而主流企业并不需要最新的模型，也不愿意为此付费。斯洛克表示，如果人工智能始终局限于整体经济中很小的一部分，其宏观经济效应将会非常有限。
人工智能模型或许会日益商品化，词元最终也可能变得足够“便宜”，从而带动营收增长和利润率扩张。但斯洛克认为，这个宏大目标实现起来可能比听起来要困难得多。他形容人工智能是一台强大的“计算器”，但同时强调，企业的运转依靠的是良好的判断力，而不是数字运算。
斯洛克在文章写道：“人工智能不理解责任、信任、文化、不成文规则、道德取舍或公平，而所有这些因素塑造着声誉风险，并最终驱动营收增长和利润率提升。”在他看来，人工智能叙事的最终走向将由“标普493”决定，而不是由七大科技巨头的财务盛景决定。
斯洛克所在的阿波罗公司管理着约1万亿美元资产，因此他对经济的判断颇具分量。他的观点也与其他警示性报告相呼应，高德纳咨询公司近期估计，到2030年底，超过3.5万亿美元的人工智能基础设施投资可能面临减值。但并非所有人都持这样的悲观看法，高盛经济学家扬·哈祖斯认为，人工智能投资最终将转化为更高的生产率和全经济范围的利润回报。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Tolsten Slocum, the chief economist of Apollo Global Management, recently wrote a sharp report by publish, stating that the real reach of the so-called “artificial intelligence revolution” is far less widespread than that of propaganda. The data show that the macroeconomic impact of artificial intelligence is still very limited, which may mean that we will have a rather interesting period in terms of financial trends and global economic trends.
According to Slock’s analysis in the Daily Spark column, despite the huge investment in recent advances in research around mega-linguistic models, and the huge wave of noise, artificial intelligence has so far not yielded any financial returns to the business community outside the United States technological industry. A small number of technology-related companies do have significant gains, but the vast majority of enterprises in the Standard 500 index do not see any results in their inputs to their generating artificial intelligence or large language models.
According to Slock, inflation-adjusted per capita revenue is rising only in the “seven giants of technology”, a group that has the highest weights in science and technology in the Standard 500 index, including Weeda, Tesla, Amazon, Microsoft, the Google parent company Alphabet, Apple and Meta. The profitability of this group is equally high. The rest of the companies on the index were described by him as “placard 493,” and their average per capita earnings and profit margins were even.
In addition, the Apollo economist did not see any signs that anyone ' s industrial intelligence was driving productivity growth. In short, the generation of artificial intelligence and large language models have no impact on the profitability of enterprises outside the technology industry. Without any tangible results, mainstream standard 500 enterprises have begun to contract the huge costs of artificial intelligence by treating terminological consumption as a cost to margins rather than as a potential source of future growth.
Venture-supported companies are increasingly spending on artificially intelligent characters, and mainstream enterprises do not need up-to-date models and are reluctant to pay for paid. Slock indicated that if artificial intelligence was always limited to a small part of the overall economy, its macroeconomic effects would be very limited.
Artificial intelligence models may become increasingly commodified, and words may eventually become sufficiently “cheap” to drive growth and profit expansion. But Slocum believes that this ambitious goal may be much harder to achieve than it sounds. He described artificial intelligence as a powerful “calculator”, but stressed that the business operated on good judgement rather than on digital calculations.
In his article, Slocum wrote: “Agent intelligence does not understand responsibility, trust, culture, unwritten rules, moral trade-offs or fairness, all of which shape reputational risks and ultimately drive growth and profit margins”. In his view, the ultimate direction of artificial intelligence narratives would be determined by the “Performance 493”, rather than by the financial fortunes of the seven major technological giants.
The company Apollo, where Shlock is based, manages approximately $1 trillion in assets, and his judgement on the economy is therefore very strong. His views echoed other warning reports, and recent estimates by Gordner Consulting suggest that over $3.5 trillion worth of AII investments could be undervalued by the end of 2030. But not everyone has such a pessimism, and Goldman Sachs economist Jan Hazus believes that investment in artificial intelligence will ultimately translate into higher productivity and economy-wide profit returns.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.