Jensen Huang's 2009 Stanford Startup Talk: Vision Is Perspective
English text is being prepared.
AYi AI Notes recommends Jensen Huang's 2009 startup talk at Stanford and has re-proofread the bilingual Chinese-English subtitles for Episode 1 (4 minutes 23 seconds) of the lecture. Jensen Huang says that so-called vision is actually perspective: in 1993, when he wanted to make 3D game chips, investors on Sand Hill Road told him to his face that there was simply no video game market. The author uses this to remind readers that, at a time when models and tools are converging, the truly uncopyable moat is the unique perspective that comes from personal experience.
Topics
View original postTo be honest, I've always felt that most people simply aren't using AI the right way. Everyone thinks that buying the most expensive membership and installing a pile of automation workflows makes them an expert, but the more they use it, the lazier their brains get. Old Huang once made a particularly clear-headed point in an interview. He said that people who truly know how to use AI never expect the model to make decisions for them; on the contrary, these people are themselves questioners with extremely high cognition. They treat AI as a lever to pry open the boundaries of their own blind spots. Think carefully about the difference between these two, and it all becomes clear. Faced with the same new industry, ordinary people toss AI a line like "write me an analysis report," get back a pile of correct but soulless nonsense, and feel smugly pleased that their efficiency is so high. People with real hands-on experience, by contrast, first put forward their own hypotheses, break down the logic and the sticking points, and force AI to follow their framework to poke holes, add counter-evidence, and simulate extreme scenarios. In the past, when we used search engines, we were used to using the shortest keywords to demand ready-made standard answers, but in the era of large models, the answers themselves have completely depreciated; the quality of the question is what is truly scarce. Put plainly, this is what the ancients called "using the Way to govern the technique," but the major precondition is that you yourself must first struggle through the mud and come up with a bit of the Way. If there isn't even a decent set of judgment coordinates in your head, then no matter how top-tier a model or Agent is placed in your hands, in the end it can only degenerate into an advanced toy. https://t.co/17KGH01Kz9
View original
说实话,我一直觉得大多数人根本没用对 AI,大家以为买个最贵的会员、装一堆自动化工作流就是高手了,结果越用脑子越懒。 老黄之前在访谈里讲过一个特别通透的观点,他说真正会用 AI 的人,从来不是指望模型替自己做决定,恰恰相反,这些人本身就是带着极高认知的提问者,他们把 AI 当成一根杠杆,去撬开自己知识盲区的边界。 你仔细品品这两者的差别,就全明白了。 同样面对一个新行业,普通人丢给 AI 一句帮我写份分析报告,拿回一堆正确但毫无灵魂的废话,心里还美滋滋觉得效率真高,而真正有实战体感的人,会先抛出自己的假设、拆解逻辑和卡点,逼着 AI 顺着他的框架去挑刺、去补充反面证据、去推演极端情况。 以前我们用搜索引擎,习惯用最短的关键词去讨要现成的标准答案,可在大模型时代,答案本身已经彻底贬值了,提问的质量才是真正的稀缺品。 说白了就是老祖宗讲的以道御术,但大前提是你自己得先在泥潭里摸爬滚打出一点道来,脑子里如果连个像样的判断坐标都没有,再顶级的模型和 Agent 塞到你手里,最后也只能沦为一个高级玩具而已。 https://t.co/17KGH01Kz9
Body
说真的,花几十万去读两年商学院,真不如把黄仁勋15 年前在斯坦福教室里的这段创业分享反复刷几遍,老黄给斯坦福学生讲出来的每一句商业直觉,放到今天看全都被一一印证了,含金量比上两年 MBA 还顶。 很多人以为商业大佬创业靠的是能预测未来的超能力,但 2009 年老黄在斯坦福讲的第一堂课就把这层滤镜撕碎了,1 993 年他刚想做 3D 游戏芯片的时候,沙丘路的投资人当面跟他说根本没有电子游戏这个市场,连他亲妈听完都劝他,你干嘛不去找份正经工作。
这期 4 分 23 秒的视频是整场经典讲座的第 1 集,我重新精校了中英双语字幕,建议每个正在找方向、或者被周围人泼冷水的兄弟都完整看三遍。
整条视频最打动我的地方,是他一上来就把远见这个高大上的词给祛魅了,他说所谓远见,其实就是视角,你看世界的方式和别人不一样,而这副眼镜每个人都可以有。
1993 年那会儿电脑还在用 Windows 3.1,世界上最快的处理器只有 66 兆赫兹,他自己都调侃说今天连球鞋里都不会塞这种芯片,
在当时的投资人和长辈眼里,电脑就是用来办公的,花重金把工作站级别的 3D 图形技术做便宜拿来打游戏,简直是不务正业。
但老黄这帮人偏偏是伴着电子游戏长大的第一代,游戏的乐趣和爆发力对他们来说是不用论证的常识,别人看不见,仅仅是因为上一代人没有这副眼镜。
说实话,放到今天做 AI、做产品、做内容也是一模一样的道理,现在大家手里的模型和工具越来越趋同,
真正抄不走的护城河,从来不是又多写了几行提示词,而是你过去的人生经历里,那些你觉得这不是明摆着吗、周围人却觉得奇怪或不重要的独特视角。
远见从来不是站在山顶俯瞰全世界的神通,而是你敢不敢认真对待自己这副不一样的眼镜,并且愿意在别人劝你找份正经工作的时候,咬着牙把路走通。