IBM's Cohn on The Next Phase of AI

以下是将内容的翻译： 对IBM副董事长 Gary Cohn 的采访围绕着人工智能的演进和应用、投资人工智能领域的挑战与机遇以及更广泛的经济议题，如关税和生产力展开。 Cohn 强调了 IBM 在人工智能领域的悠久历史，甚至早于它在 Jeopardy 智力竞赛中的获胜。他将过去依赖预先编程数据库的人工智能与今天能够访问和处理实时数据的人工智能进行了对比。他描述了人工智能从 CEO 们讨论但没有明确实施计划的概念，演变为现在的现实，即企业越来越多地识别出该技术的具体、有影响力的应用。 他强调了关注人工智能驱动的生产力提升的重要性。他通过强调 IBM 实施 HR 聊天机器人的实际应用，阐述了人工智能的用途。这个聊天机器人可以自动化执行诸如生成推荐信之类的常规任务，从而将人力资源员工从更有价值的活动中解放出来，并提高员工满意度。他还讨论了人工智能在代码辅助方面的潜力，简化软件开发，并通过代码移植帮助金融机构实现其分散的 IT 系统的现代化。 针对人工智能开发成本高昂的问题，Cohn 承认需要大量投资，特别是对于大型语言模型，但他强调了小型、目标明确的模型的潜力。他赞扬了像 DeepSeek 这样的公司，展示了可以针对特定任务定制的较小模型的效率，为许多组织提供了一种更具成本效益的方法。 讨论随后转向了人工智能领域的投资格局。 Cohn 将当前的人工智能热潮与互联网泡沫进行了类比，警告某些领域可能出现过度支出和投资回报率低的情况。然而，他认为这种投资对于构建未来的技术基础设施是必要的。 从投资者的角度来看，Cohn 承认小型、创新型人工智能公司具有吸引力，但预测拥有大量资本支出预算的大型公司可能会在长期内主导人工智能领域。他指出，大型公司通常收购有前途的初创公司，或者通过风险投资部门投资它们，从而将它们的创新整合到更大的运营中。 关于更广泛的并购格局，Cohn 指出，与年初的乐观情绪相比，交易活动有所放缓。他将此部分归因于不断变化的市场估值，也归因于关键监管官员的任命被推迟。他认为环境已经从“信任我”的方式转变为“给我看”的方式，需要为交易提供更严格的理由。 谈到人工智能对生产力和就业的影响时，Cohn 解决了对失业问题的担忧。他借鉴了诸如内燃机和互联网等技术进步的历史例子，论证了技术创新最终会通过提高生产力并开辟新的商机来创造更多就业机会。虽然人工智能可能导致技能组合的转变，但他认为它最终将人们从平凡、重复性的任务中解放出来，使他们能够专注于更有意义和更有成效的角色。 展望未来，Cohn 预测量子计算的到来会比预期快得多，可能在五年内实现。 关于人工智能监管的话题，Cohn 建议，受监管行业中的人工智能应用也应该受到监管，而那些在非监管行业中的应用应该保持不受过度监管的自由。他警告说，不要以人工智能为借口，随意过度监管或放松监管行业。 最后，谈话触及了关税问题。 Cohn 强调需要澄清征收关税的目标，质疑它们是为了保护国内产业、产生收入还是实现其他战略目标。他警告说，关税可能会对低收入消费者产生累退影响，因为它们会增加必需品的价格。 他提倡累进税制。

The interview with Gary Cohn, Vice Chairman of IBM, centered around the evolution and application of AI, the challenges and opportunities of investing in the AI space, and broader economic topics like tariffs and productivity. Cohn emphasized IBM's long history in AI, predating even the Jeopardy win. He contrasted the AI of the past, which relied on pre-programmed databases, with today's AI, which can access and process real-time data. He described the evolution of AI from a concept CEOs were discussing without clear implementation plans to a reality where businesses are increasingly identifying specific, impactful applications for the technology. He stressed the importance of focusing on productivity gains driven by AI. He illustrated the practical use of AI by highlighting IBM's implementation of an HR chat bot, which automated routine tasks like generating reference letters, freeing up HR staff for more valuable activities and improving employee satisfaction. He also discussed AI's potential in code assistance, streamlining software development, and helping financial institutions modernize their disparate IT systems through code porting. Addressing concerns about the high cost of AI development, Cohn acknowledged the substantial investments required, particularly for large language models, but he emphasized the potential of smaller, targeted models. He lauded the work of companies like DeepSeek in demonstrating the efficiency of smaller models that can be tailored for specific tasks, offering a more cost-effective approach for many organizations. The discussion then shifted to the investment landscape in the AI space. Cohn drew parallels between the current AI boom and the internet bubble, warning of potential overspending and low returns on investment in certain areas. However, he argued that such investment is necessary to build the technological infrastructure for the future. From an investor's perspective, Cohn acknowledged the allure of small, innovative AI companies but predicted that larger companies with substantial capital expenditure budgets are likely to dominate the AI landscape in the long run. He pointed out that large corporations often acquire promising startups or invest in them through venture arms, integrating their innovations into larger operations. Regarding the broader M&A landscape, Cohn noted a slowdown in deal activity compared to the initial optimism at the start of the year. He attributed this partly to changing market valuations and also to the delayed appointment of key regulatory officials. He believes the environment has shifted from a "trust me" approach to a "show me" approach, requiring more rigorous justification for deals. Moving to the impact of AI on productivity and employment, Cohn addressed concerns about job losses. He drew on historical examples of technological advancements, like the internal combustion engine and the internet, to argue that technological innovation ultimately creates more jobs by increasing productivity and opening up new business opportunities. While AI may lead to a shift in skill sets, he believes it will ultimately free people from mundane, repetitive tasks and enable them to focus on more fulfilling and productive roles. Looking ahead, Cohn predicted that quantum computing will arrive much sooner than expected, perhaps within five years. On the topic of AI regulation, Cohn suggested that AI applications in regulated industries should also be regulated, while those in non-regulated industries should remain free from excessive oversight. He cautioned against using AI as a pretext to over-regulate or deregulate industries arbitrarily. Finally, the conversation touched on tariffs. Cohn emphasized the need to clarify the objectives of imposing tariffs, questioning whether they are intended to protect domestic industries, generate revenue, or achieve other strategic goals. He warned that tariffs can have a regressive impact on low-income consumers by increasing the cost of essential goods. He advocated for a progressive tax system.