IT之家 9 月 25 日消息,在接受《纽约时报》采访时,英伟达首席执行官黄仁勋认为,当前行业和民众对 AI 的恐慌情绪已经过头了,并认为当前 AI 不一定会替代职业,但会首先替代任务。
黄仁勋把工作拆分为任务(例如写代码、整理资料和回复客户等)和目的(诊断疾病、应对工程难题、服务客户等),并认为 AI 通常先自动化具体任务,不会必然消灭整个职业。
例如,软件智能体可能大量生成代码,但软件工程师仍需要理解需要、外观平台、权衡风险、验证结果,并把工艺连接到真实难题上。
黄仁勋以工程师为例,认为 AI 可能下降写代码本身的稀缺性,但定义难题、外观架构、验证平台和承担责任的关键性会上升。
针对某些科学专家认为“AI 会在 10 年内毁灭人类”观点,黄仁勋表示:“他们的所有预测都错了,所谓的 10% 概率既没有科学依据,也没有研究兼容…… 这些预测只会增加。”IT之家附上相关截图如下:
主持人 Ezra Klein 指出历史上的自动化确实会消灭部分工作,而且 AI 比过去的工艺传播更快、适用范围更广。对此黄仁勋认为 AI 会提高生产力,并可能创造新的行业和岗位,但不等于人人都会受益。
短期看,具体的人可能先失去任务、收入或议价能力,而新机会未必马上浮现。不过从长期看,工艺可能创造新的需要和职业。不过转型速度和分配方式可能会影响到当前社会就业。
黄仁勋把 AI 比作多层结构:能源和芯片、数据中心、模型,以及最上面的应用层。他认为真正改变医院、学校、制造业和法律服务的,是应用层,而不是模型排行榜本身。
黄仁勋表示,人工智能扩张可能先带来能源代价。将来几年仍需要化石燃料支撑 AI 发展。原因在于,当前清洁发电能力尚未达到他预期的 AI 产业规模。大型数据中心持续增加,也带动电力需要快速上升。
他认为,限制 AI 数据中心建设可能拖慢实用工艺的发展。黄仁勋同时表示,AI 长期可推动能源生产投资,并协助研究人员改进电网、材料和气候平台。
本文来源:https://www.ithome.com/1/007/051.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
IT之家 reported on September 25 that, in an interview with The New York Times, NVIDIA CEO Jensen Huang believes that the current panic over AI in the industry and among the public has been overblown, and that AI may not necessarily replace jobs, but it will first replace tasks.
Huang breaks work down into tasks (such as writing code, organizing materials, and replying to customers) and purposes (diagnosing diseases, tackling engineering challenges, serving customers, etc.), and believes that AI usually automates specific tasks first and will not necessarily eliminate entire occupations.
For example, software agents may generate large amounts of code, but software engineers still need to understand requirements and platform design, weigh risks, validate results, and connect engineering to real problems.
Using engineers as an example, Huang believes that AI may reduce the scarcity of writing code itself, but the importance of defining problems, designing architecture, validating platforms, and taking responsibility will increase.
In response to some scientific experts' view that "AI will destroy humanity within 10 years," Huang said: "All their predictions have been wrong. The so-called 10% probability has neither scientific basis nor research support... These predictions will only increase." IT之家 attached the relevant screenshot below:
Host Ezra Klein pointed out that automation in history has indeed eliminated some jobs, and that AI spreads faster and applies more broadly than past technologies. In response, Huang believes that AI will increase productivity and may create new industries and jobs, but that does not mean everyone will benefit.
In the short term, specific individuals may lose tasks, income, or bargaining power first, and new opportunities may not emerge immediately. In the long term, however, technology may create new needs and occupations. Nevertheless, the pace of transformation and the distribution model may affect current social employment.
Huang compares AI to a multi-layered structure: energy and chips, data centers, models, and the application layer at the top. He believes that what will truly transform hospitals, schools, manufacturing, and legal services is the application layer, not the model leaderboards themselves.
Huang said that the expansion of AI may first bring an energy cost. In the coming years, fossil fuels will still be needed to support AI development. The reason is that current clean power generation capacity has not yet reached the scale of the AI industry he envisions. The continued increase in large data centers is also driving rapid growth in electricity demand.
He believes that restricting AI data center construction could slow the development of practical technology. Huang also said that in the long term, AI can drive investment in energy production and help researchers improve power grids, materials, and climate platforms.
