| 广告联系 | 繁体版 | 手机版 | 微信 | 微博 | 搜索:
欢迎您 游客 | 登录 | 免费注册 | 忘记了密码 | 社交账号注册或登录

首页

温哥华资讯

温哥华地产

温哥华教育

温哥华财税

新移民/招聘

黄页/二手

旅游

黄仁勋再发署名长文 系统阐述AI产业的发展逻辑


请用微信 扫一扫 扫描上面的二维码,然后点击页面右上角的 ... 图标,然后点击 发送给朋友分享到朋友圈,谢谢!



3月10日,英伟达CEO黄仁勋罕见在个人署名文章中系统阐述AI产业的发展逻辑,这是他自2016年以来发表的第七篇公开长文。

他指出,AI不应被理解为单一模型或应用,而是一个正在形成的基础设施体系。


人工智能(AI)是当今塑造世界的最强大力量之一。它不仅是一个聪明的应用程序或单一的模型;它是如同电力和互联网一样不可或缺的基础设施。

在他看来,AI产业正在经历一次类似工业革命级别的技术基础设施建设。当前全球已投入数千亿美元,但整体建设仍处早期阶段。

黄仁勋表示,AI是“五层蛋糕”基础设施——能源、芯片、基础设施、模型、应用,还需数万亿美元建设。



他预判,未来几年,传统的软件和APP形态或将消失,一种全新的软件范式AI Agent(智能体)极有可能成为主流。

针对AI发展带来的就业担忧,黄仁勋认为,AI非但不会削减岗位,反而会创造大量新的就业机会,尤其是在基础设施和熟练技术工种领域,支持AI基础设施建设所需的劳动力极其庞大,AI工厂需要电工、水管工、钢铁工人、网络技术人员、安装工和操作员等,这些都是高技能、高薪酬的岗位,且目前供不应求。

AI正在从“软件”变成实时生成智能


黄仁勋首先解释了AI与传统软件的根本差异。

过去几十年,软件基本是“预先写好的程序”。开发者编写算法,计算机按规则执行。数据必须结构化,并通过数据库查询调用。而AI改变了这一模式。

黄仁勋写道:“这是计算机历史上第一次,机器可以理解非结构化信息——图像、文本、声音,并理解其中的意义。”


更重要的是,AI并不是从数据库中读取答案,而是 实时生成智能。

他解释:“每一个回答都是新生成的,每一次输出都取决于上下文。计算机不再只是执行指令,而是在推理。”

由于智能是在实时生成,这迫使整个计算架构重新设计。

AI产业的“五层结构”

在文章中,黄仁勋提出了一个AI产业的结构框架:五层技术栈——能源、芯片、基础设施、模型、应用。


1     已经有 8 人参与评论了, 我也来说几句吧
上一页1234下一页
注:
  • 新闻来源于其它媒体,内容不代表本站立场!
  • 在此页阅读全文
    猜您喜欢:
    您可能也喜欢:
    共有 8 人参与评论    (其它新闻评论)
    评论1 游客 [辛.香.氏.说] 2026-03-11 08:21
    At the end of the day, all those AI investments must be paid back with profits by their users. At the beginning, they allow free rides or free trials to make users addicted. Then they will start to charge for their services. They will charge a lot when their services become essential parts of businesses or personal lives. That is what happened for all high tech services.
    评论2 游客 [辛.香.氏.说] 2026-03-11 08:30
    More than four decades ago, lots of us used BASIC language to create short programs to solve simple problems on DOS operating systems. They were isolated PCs. When simulation models were developed, we used powerful super computers. It was all about how fast to do a single operation by computers. Now we have AI which needs much more powerful capacities to do a single operation. It can solve lots of general problems but is not always cost-efficient for all applications.
    评论3 游客 [辛.香.氏.说] 2026-03-11 08:39
    Before the term of AI become popular, technologies have already been developed to solve lots of problems to help increase the efficiency in lots of industries by using pre-programmed software or systems with their industry specific database. They are not applicable for general purposes. Therefore, they are still relatively cheaper than AI which can generate new information by reasoning. Reasoning models demand super high computing capacities which are supported by high energy, high performanc...  [更多...]
    评论4 游客 [辛.香.氏.说] 2026-03-11 08:50
    AI at the stage is just technical explorations and venture investments. Lots of leading AI companies put heavy investments on the developments for both models and infrastructures to bet future monopoly. AI is still at its early stage. It may have lots of paths and may not be only one way for all. General AI will not take all areas but part of general services. Specialty AI might be more cost-efficient in the rest of areas where not all other knowledge is required.
    上一页1下一页
    我来说两句:
    评论:
    安全校验码:
    请在此处输入图片中的数字
    The Captcha image
    Terms & Conditions    Privacy Policy    Political ADs    Activities Agreement    Contact Us    Sitemap    

    加西网为北美中文网传媒集团旗下网站

    页面生成: 0.0256 秒 and 6 DB Queries in 0.0036 秒