新智元报道
如果说，当今世界有人能看清人类未来十年的走向，Dario Amodei绝对是其中之一。
他曾接受了一场非常深入的访谈。在这场对话中，他坦诚分享了自己对未来的预测——
「AI海啸正向我们扑来，近在咫尺。但整个社会却假装那只是海市蜃楼。」
未来十年，什么行业会被彻底颠覆？
25岁的年轻人今天应该学什么专业？
编程是不是真的穷途末路？普通人如何在这场大洗牌中保住饭碗？
访谈中，主持人问了一个所有人都无比焦虑的问题：
如果我今天25岁，想在未来十年实现财富上的成功，我应该去学什么？编程还有用吗？什么行业还有红利？
Dario的回答很残酷。
他直言不讳地说：「我认为，Coding会最先消失，或者说，最先被AI模型完全取代。」
在过去二十年里，「转码」几乎是普通人跨越阶层的最快捷径。
无数年轻人涌入CS专业，只为在大厂谋得一份高薪的「码农」工作。
但Dario无情地指出，纯粹的敲代码时代，正在终结！在Anthropic内部，Claude Code已经在指数级进化。
但是，请注意，Dario并没有判软件行业的死刑。
他提出了一个极其关键的区分：Coding（编码） ≠ Software Engineering（软件工程）。
写出具体的代码行会被AI代劳，但「软件工程」——即作为一个架构师，去思考如何设计一个系统、如何满足用户的真实需求、如何管理由多个AI代理组成的团队——这项工作会存在更长时间。
「这是一种极其强大的『比较优势』。」Dario解释道。
Dario 给出了一个极具颠覆性的数学公式：5% + 95% = 20倍生产力。
人类的比较优势并没有消失，只是发生了剧烈转移。在未来的工作流中，AI 将承担 95% 的执行工作。而人类，只需死死守住那最核心的 5%。
这 5% 是什么？
是判断力。是品味。是「知道我们要去哪儿」。
这意味着，你的生产力将被放大20倍！未来不需要一屋子的初级程序员，只需要一个懂业务、懂人性的「超级个体」，带着一群不知疲倦的AI，就能干翻一个传统的百人团队。
Dario在访谈中用了一个比喻：海啸。
令我惊讶的是，目前的AI模型已经如此接近人类智能的水平，但整个社会似乎完全没有意识到即将发生什么。
这就像一场海啸向我们袭来，它已经在海平线上清晰可见，近在咫尺，但人们却在疯狂地寻找借口，安慰自己说：『那不是海啸，那只是光影的折射。』
为什么Dario对AI的进化如此笃定？因为他是最早发现并笃信「Scaling Law」的人之一。
对于Scaling Law，Dario用了一个极具画面感的比喻：就像一场化学反应。
如果你想生火，你需要不同的配料。如果你把「数据」和「算力」按照正确的比例放入「模型规模」的反应炉中，这场化学反应的最终产物，就是「智能」 。
这不是玄学，而是残酷的物理和数学规律。只要算力和数据继续增加，AI就会变得越来越聪明。
5年前，电脑连一张像样的猫咪图片都画不出来；而今天，你可以给Claude发一段猴子杂耍的视频，问它「球在猴子手里换了多少次」，它能精准地回答你。
这就是Dario感到的「急迫感」。人们总是习惯用线性的思维去预测未来，认为「这个改变太大了，不可能发生」，但技术正在以指数级的速度狂飙。
当海啸来临时，如果你还站在海滩上捡贝壳，结局注定是毁灭。
既然「智力劳动」正在被AI疯狂吞噬，那么我们到底该学什么？我们的孩子该选什么专业？
Dario给出了一份「未来生存指南」。他认为，未来的护城河，深深地扎根于以下三个领域：
第一条退路：极度「以人类为中心」的工作
AI可以写出完美的商业计划书，但AI无法在酒桌上和客户建立信任；AI可以给出最精准的医疗诊断，但AI无法握住患者的手。
Dario提到了放射科医生的例子。曾经有人预测AI会淘汰放射科医生，因为AI看片子比人准多了。
但现实是，放射科医生并没有失业，因为他们工作中最具技术含量的「看片子」被AI替代了，但他们现在把时间花在「与患者沟通、解释病情、安抚情绪」上。
未来的高薪职业，将属于那些能够链接人、理解人、调动人的角色。情商、共情能力、以及建立复杂社会关系网络的能力，将成为最昂贵的稀缺品。
第二条退路：深入「物理世界」
无论AI在虚拟的数字世界里多么手眼通天，它目前依然很难在现实世界里帮你通马桶、拧螺丝。
Dario强烈建议，年轻人应该关注那些「物理世界」与「传统工程」交汇的领域，比如半导体制造、新能源基建、高端制造业。
当软件世界被AI卷成红海时，物理世界的建设反而因为其「笨重」、「需要真实触感」而成为避风港。如果能把AI的分析能力与物理世界结合起来，你将天下无敌。
核心底层能力：街头智慧与批判性思维
在一个AI可以随时生成假视频、假图片、假新闻的世界里，什么能力最重要？
Dario认为，是批判性思维和街头智慧。
「你不希望被假象蒙蔽，你不希望拥有错误的信念，你更不希望被诈骗。」
在未来，分辨「真实与虚假」将成为生存的基本功。那些只会死记硬背、缺乏独立思考能力的人，将成为AI时代的第一批韭菜。
在这场长达数小时的访谈中，主持人向Dario抛出了非常刁钻的问题——
「如果现在给你100块钱，只能买一支股票或者押注一个行业（除了你自己的AI赛道），你会买什么？」
Dario笑了。他没有推荐微软，也没有推荐英伟达。
他给出了一个令人意外，但又在情理之中的答案：生物技术。
不要忘了，在创立估值400亿美金的AI帝国之前，Dario原本是一个顶尖的物理学和生物物理学博士，曾在斯坦福医学院做博后。
他最初搞AI，就是因为觉得生物系统的蛋白质结构太复杂了，人类脑容量不够，需要发明AI来帮他治病。
Dario断言：「生物技术即将迎来一场由AI驱动的文艺复兴。我们即将治愈大量令人绝望的疾病。」
他特别指出了一个细分赛道：多肽疗法和基于细胞的疗法（如CAR-T）。
为什么？因为传统的化学小分子药物，设计起来就像碰运气；而多肽和mRNA等新一代生物技术，具有「可编程」的特性。它们就像是生物界的计算机代码，可以通过AI进行数字化计算和连续优化。
当AI的算力与「可编程的生物学」相撞，人类在对抗癌症、衰老和绝症的战役中，将迎来重大突破。
如果你在寻找长期的投资风口，或者孩子正在犹豫要不要读生物医学，Dario的这个判断，值得参考。
随着AI接管越来越多的认知工作，很多人恐惧：人类是不是正在把自己的大脑「外包」出去？我们会不会像《机器人总动员》里那样，变成连算数都不会的傻子？
Dario坦言：如果我们粗心大意地使用AI，人类确实会变笨。
他们研究发现，如果学生只是让AI帮忙写作业、或者程序员只是盲复制AI的代码，他们的技能确实在退化。
但是，选择权在人类自己手里。就像计算器发明后，人类并没有丧失数学能力，反而利用计算器算出了登月轨道的参数。Dario自己现在依然经常在脑子里做心算，因为「这能让数学与我的思维过程深度融合」。
AI永远比你聪明，但不代表你不需要学习。学习的过程，是重塑你大脑神经元的过程。
对于想要在AI时代创业或做超级个体的普通人，Dario给出了以下建议——
1.绝对不要做「套壳产品」。
如果你只是简单地调一下Claude或ChatGPT的API，做个漂亮的UI界面，加点提示词就拿去卖钱。Dario警告：「这种东西毫无护城河可言，任何人都能吞噬你的收入，不用等AI巨头出手，你自己就会先死掉。」
2.建立极深的垂直壁垒。
Anthropic绝不会去碰那些高度专业化的垂直领域，比如生物医药跨界AI、高度监管的金融服务业等。巨头看不上、或者懒得干的脏活累活和垂直Know-how，就是普通创业者的生存土壤。
3.把「提示词工程」当成弹钢琴来练。
很多人抱怨AI不好用，是因为他们根本不会用。Dario引用了一句名言：「写提示词就像弹钢琴。你不能指望一坐到琴凳上，就能弹出贝多芬。」
为了让非技术人员也能用好AI，Anthropic开发了无需写代码的Claude Co-work。但无论工具多简单，你都需要去花时间「实操」。
我们需要去试错，去给AI设定上下文，去建立AI智能体机制。谁先学会掌握这台「蒸汽机」，谁就能在大洗牌中拿到头等舱的船票。
在这场深度访谈的最后，Dario说了这样一段话：
在过去10年里，我反复经历这样一种体验：当你根据简单的曲线外推，或者用简单的逻辑去推演未来时，你会得出一些反直觉的结论。很多人会说：『那太疯狂了，不可能发生。』
但最终，它真的发生了。你几乎可以『免费』地预测未来，只需要几个经验观察，加上第一性原理的思考。」
AI时代的第一性原理，就是智力成本将无限趋近于零。任何可以被逻辑化、数据化、重复化的脑力劳动，都将一文不值。
人类能做的，就是去理解人心，去连接物理世界，去掌控AI，去做那5%能掌舵的人。
参考资料：
编辑：Aeneas 大卫
不喜欢
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。
New Ji-won.
Dario Amodei is definitely one of those who can see the course of humanity for the next decade in today ' s world.
He had a very in-depth interview. In this conversation, he frankly shared his vision of the future -
"Ai tsunami is coming at us, near. But the whole society pretended that it was just a mirage."
What industry will be completely subverted in the next decade?
What should a 25-year-old learn today?
Is programming really the end of the road? How can ordinary people keep their jobs in this big shuffle?
During the interview, the moderator asked a question that everyone had been extremely anxious:
What am I supposed to learn if I'm 25 today and want to succeed in wealth over the next 10 years? Is programming working? What kind of business is there?
Dario's answer was cruel.
And he said, "I think Coding will be the first to disappear, or be completely replaced by the AI model."
In the last two decades, "turn code" has been almost the fastest route for ordinary people to cross class.
Numerous young people have entered the CS specialty, only to find a well-paid "cod farmers" in large factories.
But Dario is relentlessly pointing out that the era of pure code knocks is coming to an end! Within Anthropic, Claude Code has evolved at the index level.
However, note that Dario has not sentenced to death in the software industry.
He drew a crucial distinction: Coding (coding) Software Engineering.
Writing specific code lines will be done by AI, but "software engineering" — that is, as an architect, thinking about how to design a system, how to meet the real needs of users, and how to manage a team of multiple AI agents — will last longer.
"It's an extremely powerful comparative advantage." Dario explains:
Dario gave a very subversive mathematical formula: 5% + 95% = 20 times productivity.
The comparative advantages of humankind have not disappeared, but rather have shifted dramatically. In future workflow, AI will undertake 95% implementation. And humans, just hang on to that core five percent.
This 5% is what?
It's judgment. Taste. It's "know where we're going."
That means your productivity will be increased 20 times! The future does not require a house full of junior programmers, but a business- and humanistic "superman" with a bunch of tired AIs, and a traditional 100-person team.
Dario used a metaphor in his interview: tsunami.
I am surprised that the current AI model is so close to human intelligence, but society as a whole seems completely unaware of what is going to happen.
It was like a tsunami that came to us, and it was visible on the horizon, and it was close, but people were searching madly for excuses and comforting themselves by saying, "It was not a tsunami, it was a reflection of light."
Why is Dario so sure about the evolution of AI? Because he was one of the first people to find out and believe in Scaling Law.
For Scaling Law, Dario uses a very graphic metaphor: it's like a chemical reaction.
If you want a fire, you need different ingredients. If you place "data" and "calculation" in the reactor of "model scale" at the right rate, the end product of this chemical reaction is "smart".
It's not encyclopedia, it's cruel physics and mathematics. As long as the numbers continue to grow, AI becomes smarter.
Five years ago, computers couldn't even paint a decent picture of a cat; and today, you can send Claude a video of monkey juggling and ask it, "How many times has the ball changed in the monkey's hand," and it answers you with precision.
This is what Dario feels like. People are used to project the future with linear thinking that "this change is too big to happen," but technology is raging at an exponential rate.
And when the tsunami comes, if you stand on the beach to collect shells, the end is doomed.
Now that "intellectual labor" is being consumed by AI's madness, what exactly should we learn? What profession should our children choose?
Dario gave a "guideline for future survival". In his view, the future moat is deeply rooted in three areas:
First retreat: the extreme "human-centred work"
AI could write a perfect business plan, but AI could not build trust at the wine table and among clients; AI could give the most accurate medical diagnosis, but AI could not hold the patient ' s hand.
Dario referred to the example of a radiologist. There were predictions that AI would eliminate radiologists because it was better than humans to watch.
But the reality is that radiologists are not unemployed, because the most skilled "watches" in their work have been replaced by AI, but they now spend their time "communicating with the patient, explaining his condition, calming emotions."
Future high-paying occupations will belong to those who can link people, understand people and move people. Intriguing, interoperable capabilities and the ability to build complex networks of social relationships will become the most expensive and scarce.
Second exit: deep into the "physical world"
No matter how wide-eyed AI is in the virtual digital world, it is still difficult to help you with toilet and screw in the real world.
Dario strongly recommends that young people focus on areas where the “physical world” intersects with “traditional engineering”, such as semiconductor manufacturing, new energy infrastructure, and high-end manufacturing.
When the software world was turned into the Red Sea by AI, the physical world was built as a safe haven because of its "dickness" and "need real touch". If you combine the analytical capabilities of AI with the physical world, you will be invincible.
Core bottom competencies: street wisdom and critical thinking
What is the most important ability in a world where AI can generate fake videos, false pictures, false news at any time?
Dario believes it is critical thinking and street wisdom.
"You do not want to be blinded, you do not want to have the wrong faith, and you do not want to be deceived."
In the future, the identification of "real and false" will be an essential function of survival. Those who are only hard-headed and lack the ability to think independently will be the first bellies of the AI era.
In this hour-long interview, the host threw out a very bad question for Dario...
"What would you buy if you gave you $100 right now, only one stock or a bet on one business?"
Dario laughed. He didn't recommend Microsoft, he didn't recommend Inverda.
He gave an unexpected but sensible answer: biotechnology.
Don't forget, before the creation of the 40 billion-dollar AI empire, Dario was a top doctorate in physics and biophysics, after doing his job at Stanford Medical School.
He started with AI because he felt that the protein structure of the biological system was too complex and that human brain capacity was insufficient and that it needed to be invented to cure him.
Dario asserted: "Biotech is about to usher in an AI-driven Renaissance." We are about to heal a large number of desperate diseases.”
In particular, he noted a break-down track: polypesal therapy and cell-based therapy (e.g. CAR-T).
Why? Because of traditional small chemical molecules, they are designed like luck; and new generations of biotechnologies, such as polyazine and mRNA, have a "programmable" character. They're like biological computer codes that can be digitized and continuously optimized through AI.
When AI's calculations are colliding with "programmable biology", human beings will make a major breakthrough in the battle against cancer, ageing and terminal illness.
If you're looking for a long-term investment, or if the kid's hesitating to go to biomedicine, Dario's judgment is worth mentioning.
As AI takes over more and more cognitive work, many fear that humans are "outsourcing" their brains. Are we going to be the same as those robotics who can't even count?
Dario honestly said that if we use AI carelessly, humans will really be stupid.
They have found that if students simply ask AI to help with their homework, or programmers simply blindly copy AI codes, their skills are indeed degraded.
But the choice is in the hands of humanity itself. Just as humans did not lose their mathematical ability after the calculator was invented, they used the calculator to calculate the parameters of the moon orbit. Dario's own mind is still in his head, because "it's a way to blend math into my mind."
AI's always smarter than you, but it doesn't mean you don't have to study. The learning process is the process of reshaping your brain neurons.
Dario gave the following suggestions to ordinary people who wanted to start a business or become super individuals in the AI era:
No "shell product" should be made.
If you just switch Claude or ChatGPT's API, do a nice UI interface, and add a little prompt, you sell it. Dario warned: "There is no moat of this, and no one can devour your income, and you will die first, without waiting for the giant al-Ai."
2. Establish deep vertical barriers.
Anthropic would never touch on highly specialized vertical areas such as biomedicine cross-border AI, highly regulated financial services, etc. A giant who does not look like, or is lazyly engaged in, dirty work and vertical Know-how is the living ground for ordinary entrepreneurs.
Use Project prompt as a piano.
A lot of people complain about AI because they can't use it. Dario quotes a famous phrase: "Write prompt is like playing the piano." You can't expect Beethoven to come out of a chair."
In order for non-technicals to be able to use AI, Anthropic developed Claude Co-work. But no matter how simple the tools are, you need to spend time on "acting."
We need to try and set AI up context and set up the AZAI agent mechanism. Whoever learns to master the steam engine first gets the first-class tickets in the shuffle.
At the end of this in-depth interview, Dario said:
Over the past 10 years, I have experienced this experience repeatedly: When you push out of a simple curve, or you push out the future with a simple logic, you come to some counterintuitive conclusions. A lot of people would say, "That's crazy, it can't happen."
But eventually, it really happened. You can predict the future almost `free', but only a few empirical observations, plus the first theory."
The first principle of the AI era is that intellectual costs will be endlessly close to zero. Any intellectual effort that can be logically, digitally and duplicated will be worthless.
What humans can do is to understand people's hearts, to connect to the physical world, to control AI, to be the 5% who can steer.
References:
Editor: Aeneas David
I don't like it.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.