Summary of Key Points
This interview features a sober analysis of AI trends by tech analyst Benedict Evans: AI is a revolutionary technology akin to the internet, but it is currently at the “1997 internet” stage—everyone recognizes its importance, yet we haven’t fully understood its potential for disruption. While AI will replace routine tasks, high-level judgment skills (such as consulting and legal negotiations) will become even more valuable. The key to competition lies not in the AI technology itself, but in “where users deploy AI” (i.e., the distribution channels). Jobs won’t disappear, but they will be restructured; individuals need to learn about AI with real-world problems in mind. Public backlash against AI is not a bad sign; rather, it indicates that the technology is beginning to integrate into society.
I. AI is Like the Internet in 1997: Busy, but Not Yet Disruptive
Benedict compares AI to the internet in 1997: everyone was talking about building websites, but Google hadn’t yet been founded, the iPhone was still a decade away, and broadband wasn’t widely available; truly transformative applications (like e-commerce and social media) were nowhere in sight. The same is true for AI today—automated customer service, code writing, content generation, etc., are all basic uses.
He argues that truly disruptive AI applications (those that could revolutionize industries in ways we can’t yet imagine) will take 5-10 years to emerge. What surprises him the most is the rapid pace of AI’s development: just a few years ago, people were asking if AI could write articles; now, the focus has shifted from “can AI do it” to “how can we use AI to write good articles”—this shift has happened much faster than expected.
II. Counterintuitive! Will High-End Advisors Become More Valuable with AI?
You might think AI will take away jobs from advisors, but Benedict finds that the demand for high-end lawyers and strategists is actually increasing. The reason is simple: AI makes it easier to acquire knowledge (e.g., generating contracts, finding legal precedents), but judging how to apply that knowledge becomes more challenging. For example, while AI can draft contracts, it cannot understand complex negotiations, interpret ambiguous regulations, or outsmart opposing lawyers.
Moreover, as basic tasks are automated by AI, the demand for specialized expertise grows. Knowledge itself becomes less valuable; instead, the ability to interpret and apply that knowledge becomes more crucial.
III. The Real Barrier in the AI Era: Not Technology, but How You Use It
Many believe that the competition in AI lies in model capabilities, but Benedict argues that the gap between models is narrowing, and API prices are decreasing, making it difficult to win solely on technology. The real barrier is distribution—where users access AI.
Companies like Microsoft (with Copilot in Word/Excel/Teams), Google (in search), and Apple (with their AI assistants) have integrated AI into everyday tools. Users will naturally prefer these built-in features over downloading separate apps. New startups should not focus on creating better models than GPT-4; instead, they should identify specific use cases (e.g., pain points in a particular industry) and solve problems with AI, using proprietary data and customer relationships to gain an advantage. For instance, an AI that helps doctors draft medical records is more targeted than general-purpose AI.
IV. Will AI Take Your Job? Don’t Panic; It Takes Tasks, Not Jobs
Don’t worry about AI taking your job directly; it replaces specific tasks, not entire jobs. Every job consists of a combination of tasks. For example, a radiologist’s tasks include analyzing images, writing reports, and communicating with patients. AI can handle image analysis and report writing, but communication and training new doctors still require human skills.
Historically, every technological revolution (electrification, the internet) led to concerns about job loss, but demand eventually expanded, creating new opportunities. Similarly, AI may reduce the cost of legal services, making more people able to afford lawyers, leading to growth in the legal industry. For individuals, the approach is to find a problem you care about and use AI to solve it—using AI to create drafts, for example, is more useful than simply learning AI tools.
V. The More Anti-AI Sentiment, the More Important AI Is?
There’s much backlash against AI (e.g., concerns about privacy and job loss), but Benedict sees this as a good sign: when technology makes people uncomfortable, it means it has crossed from the tech world into everyday life. The internet also faced opposition (privacy issues, fake news), yet that only demonstrated its importance.
The future of AI will not be entirely positive or negative; just like the internet, it brings both benefits and challenges (connectivity and misinformation). His advice is to avoid jumping to conclusions and maintain a curious attitude, focusing on what you can do now.
In conclusion, Benedict shares two key principles: “It depends” (anything about the future is subject to context) and “it’ll probably be okay” (things usually don’t go too badly). This reflects his calm, pragmatic approach to analyzing AI trends.