Author: Techub News Compilation

Introduction

On the main stage of Salesforce's annual event Dreamforce 2025, Salesforce Chairman and CEO Marc Benioff engaged in an in-depth dialogue with Dario Amodei, Co-founder and CEO of Anthropic. As one of the most prominent companies in the current AI wave, Anthropic is known for its enterprise-level AI model Claude and its strong focus on AI safety. This conversation took place at a time when Anthropic's business is rapidly expanding and its valuation is soaring, during which Dario Amodei rarely reflected on his journey from neuroscience research to AI, leaving OpenAI to found Anthropic, and shared the latest thoughts on the pace of AI technological advancement, its future impact, and Anthropic's unique development path.

Summary

• The trajectory of AI development aligns with the early predictions of the "law of scaling," but the speed and breadth of its transformative impact in reality remain astonishing.
• Anthropic firmly focuses on the enterprise market, believing that the values of "responsibility" and "trust" are highly aligned with the needs of enterprise services, which is a key differentiating factor for its success.
• The transition from "co-pilot" to "agent" represents the most exciting technological breakthrough, as AI evolves from an assisting tool to a partner capable of performing complex tasks end-to-end.
• The relationship between AI and humans should be more "complementary" rather than simply "substitutive," which will reshape workflows in the short term but may also lead to broader and faster disruptions in the labor market in the coming years.
• Anthropic's long-term vision is to place AI at the core of every enterprise, creating immense social value by addressing diseases, empowering industries, and maintaining a commitment to reliable and responsible product promises.

From Neuroscience to AI: Starting with an Awe of "Complexity"

Dario Amodei's career began with a deep interest in life sciences. During his postdoctoral research in neuroscience and computational biology at Stanford University, he felt that the complexity of biological systems "exceeds the scale of human understanding." Faced with challenges such as diseases like cancer, he once felt a sense of despair about the pace at which humanity was advancing research.

A turning point came with the dawn of the deep learning revolution. Early deep neural network research from Google and the University of Toronto opened his eyes. "I replicated some results and thought, 'Wow, this stuff really works, it’s going to change everything,'" Dario Amodei recalled. This prompted him to decisively shift towards the field of AI, first joining Google and then OpenAI, becoming one of the earliest researchers to systematically study the "law of scaling" in AI technology.

During his time at OpenAI, he and the future co-founders of Anthropic clearly saw two trajectories: first, the capabilities of AI would exhibit exponential growth and reshape the global economy within a few years; second, these AI systems are fundamentally different from traditional software. "Building traditional software is like constructing a skyscraper; you need to design a blueprint and every detail. AI models are more like nurturing plants or baking a cake; you set a high-level 'recipe,' but the result has a degree of unpredictability," Dario Amodei explained. This unpredictability, combined with the tremendous economic impact that AI is about to produce, convinced him that "safety" and "responsibility" are crucial. Based on the dual belief in technological potential and the necessity for governance, he and seven other co-founders left OpenAI around 2020 to jointly establish Anthropic.

"Responsibility" and "Trust": Natural Allies in the Enterprise Market

In response to Marc Benioff's probing, Dario Amodei specifically mentioned the unique story of leading the company together with his sister Daniela Amodei. The two grew up in San Francisco and early on shared the aspiration of "doing something big together and making a positive impact on the world." Daniela is responsible for the daily operations, organization, and business processes of the company, while Dario focuses on strategy, research direction, and company vision. This division of labor allows them to maintain focus and efficiency in their respective "operational mode" and "vision mode."

Since its founding four and a half years ago, Anthropic has experienced rapid growth. Dario Amodei revealed that the company currently holds a significant share in the number of enterprise clients and revenue, possibly even a majority. He emphasized that this is closely tied to the values of "responsibility" and "trust" that Anthropic has upheld since its inception. "Running an enterprise-focused company, you realize that these values are synergistic with serving enterprises. These values help us better serve enterprise clients, and the business model of serving enterprises, in turn, reinforces our values." He believes that when enterprises entrust critical business processes to AI, predictability, reliability, and avoidance of unexpected harmful outputs are essential needs, and this is precisely the moat that Anthropic strives to build.

Marc Benioff also confirmed that Salesforce is an early investor in Anthropic and that a close partnership has been established, with the Claude model deeply integrated into products such as Slack.

The Rise of Agents and the Evolution of Human Roles

When discussing the most exciting technological advances today, Dario Amodei pointed directly to the breakthrough in "agent capabilities." AI is transitioning from a tool that passively responds to commands to an "agent" capable of proactively executing complex tasks end-to-end. He cited coding as the most evident area currently manifesting this change. "Programmers are early adopters, and they’ve given us a glimpse into the future."

He shared a specific case: while training the latest version of the Claude model, a tricky failure occurred in the compute cluster, and engineers spent days unable to resolve it. Eventually, the team had Claude "explore the cluster itself to see if it could find the problem," and Claude identified a hidden bug that the engineers had overlooked. "It felt like having an exceptionally talented teammate," Dario Amodei said.

He even boldly predicted that in both Anthropic and many partner companies, 90% of the code is currently written by Claude. However, he immediately clarified that this does not mean a reduction in the number of software engineers or their replacement. He referred to the principle of "comparative advantage" in economics to explain: when AI takes on most of the foundational and repetitive coding work, human engineers can be freed up to focus on supervising, editing, addressing the most challenging 10% of problems, or managing a group of AI models. "This often means you need the same amount or even more software engineers because they can now be greatly empowered, with productivity potentially increasing tenfold." Dario Amodei emphasized that at this stage, this is more of a "rebalancing" of work rather than "replacement."

However, he expressed cautious concerns about the longer-term future. "I optimistically believe that 'complementarity' will continue in the short term. But I do worry that within the next two to five years, there will be a large enough economic disruption that will lead to a quicker and broader chaotic labor market than previous technological changes." He noted that as AI capabilities continue to enhance rapidly, the collaboration model between humans and AI will also evolve, possibly shifting from close collaboration to more distant supervision, which requires strong adaptability from both society and individuals.

Opportunities and Challenges in Highly Regulated Industries: The Case of Healthcare

Naturally, the conversation turned to highly regulated and high-risk industries such as healthcare, which is also a key area of collaboration between Anthropic and Salesforce. Dario Amodei acknowledged that current AI models do not achieve 100% accuracy in the medical field and should not be blindly trusted. He likened the role of models to that of a "second opinion." "Just like a doctor tells me something, I respect that they are experts, but I won’t take it as absolute fact. They are usually right, but not always."

He shared a personal experience: his sister Daniela contracted a virus during her pregnancy (which suppresses the immune system), but the doctors initially overlooked the possibility of a bacterial infection. Daniela uploaded her data to Claude, which systematically listed various possibilities, including the bacterial infection that the doctors had missed. "Claude excels at asking, 'Have you considered this? Have you thought about that?' It has this assisting, augmenting capability."

This case highlights the value of AI in information sorting and suggesting possibilities, but it also warns against the risks of blindly following AI recommendations. Dario Amodei believes that in fields like healthcare and radiology, clarifying the roles of humans and models and establishing the best collaboration model still requires extensive exploration and practical adaptation. "Programmers have already found the best practices for using Claude to write code through blog posts in a matter of months. I believe a similar process is happening in healthcare and radiology, but due to the high-risk nature of these industries and numerous barriers, this process is much slower."

Looking Ahead to 2030: Placing AI at the Core of Every Enterprise

When Marc Benioff asked about Anthropic's vision for the next five years, Dario Amodei returned to his personal passion: using AI to tackle diseases. He mentioned an article he wrote titled "Loving Machines," which depicts the immense potential of AI in curing stubborn ailments like cancer and Alzheimer's disease. "I believe AI is the right answer to solving these problems."

Beyond healthcare, he is also excited about all enterprise-level applications in finance, insurance, and beyond. He likened "enterprise" itself to a type of "super intelligence"—an entity with strategic insights, knowledge, and capabilities far exceeding individual human beings. "Even if these models become smarter than any single human in the future, there remains a significantly higher ceiling in accomplishing what a company aims to achieve in the world." Therefore, he believes that serving the enterprise market has an extremely high technical ceiling and market potential.

He confirmed recent reports that Anthropic's annual recurring revenue is nearing $7 billion and noted that the company's revenue has maintained a growth rate of tenfold each year for several consecutive years. "Of course, this pace cannot be sustained indefinitely, but the potential is enormous."

However, he then pivoted to emphasize that more important than commercial success, Anthropic hopes to be known as a company that is "trustworthy" and "responsible." "We want to be that kind of company: if you build based on our products, you won’t encounter any 'surprises,' crazy things won't happen, and the model won't speak nonsense. We believe this is crucial for us to be good stewards of this technology." This statement is seen as a nuanced yet clear delineation from certain unpredictable or controversial behaviors in the industry.

Computing Power, Partnerships, and Sustainable Growth

When discussing the current global trend of building data centers and competing for chips, Dario Amodei acknowledged that according to the law of scaling, training smarter models and serving more customers requires enormous computational support, and in the future, gigawatt-level power might be needed. Anthropic is actively seeking stable computational and energy supplies worldwide.

He also expressed concerns about potential "inflation" in certain data center transactions on the market (e.g., "double counting"), but he believes that at the end of the day, a healthy business model is fundamental. "I think the most important thing is making money. If there is demand, and customers are willing to pay, then you are willing to pay your computing suppliers, and ultimately you can obtain the computational power you need." He proudly pointed out that Anthropic is currently the fastest-growing company in this field, with the fastest revenue growth, and strong customer demand is the foundation for its ongoing investment in computational infrastructure.

Finally, Dario Amodei reiterated his pride as a local San Franciscan and his commitment to Anthropic's roots here, along with partners including Salesforce, to shape the future of AI together. The entire dialogue paints a picture of a thoughtful AI leader: one who is acutely aware of the exponential power of technology and passionately pursues it, while remaining highly vigilant about the broad social responsibilities that come with it and attempting to find a unique path of balanced development between the pragmatic demands of the enterprise market and long-term visions.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。