Summary of Key Points
In this interview, Sam Altman did not overemphasize how powerful AI will be in the future; instead, he focused on the practical obstacles and strategic choices surrounding the implementation of AI. He admitted that he had underestimated the inertia of the economy and people (technology moves fast, but corporate processes and personal habits change slowly). He pointed out that AI products are still missing the “iPhone moment” – a breakthrough in user experience. Altman stressed that computing power has become the core battleground of AI competition, where capital, energy, and supply chains are more crucial than algorithms. He made it clear that OpenAI aims to be a platform rather than a company that offers a range of products, abandoning projects like Sora to focus on general intelligence and APIs. He also discussed the dilemma of AI security, expressing concerns about both loss of control and concentration of power, and suggested solving these issues through real-world iterations. Overall, his message is one of technological optimism combined with a pragmatic approach: AI will change the world, but it will take time, and people need to adapt gradually.
Detailed Analysis
1. Why is AI not more widespread? It’s not that the technology is inadequate; it’s that people and the economy are too “lazy”
Altman’s biggest surprise was that, despite the release of GPT-4 in 2023, the software industry did not quickly transform. The reason is simple: both the economy and people have inertia:
- At the corporate level: Procurement processes, organizational structures, and data security concerns prevent companies from immediately adopting new methods.
- At the individual level: Even with tools like Codex that can automate tasks, Altman still prefers to manually handle things, just as he has for the past 20 years. He calls this “behavioral inertia,” which better explains the “AI productivity paradox” – despite technological progress, efficiency has not significantly improved.
For example, even though Netflix can deliver DVDs by mail, many people still rent movies from Blockbuster because of established habits.
2. AI hasn’t reached the “iPhone moment”: The technology is there, but the experience is awkward
Altman compares current AI to the Palm Treo before the iPhone: the technical components are in place, but the user experience has not improved. AI tools are like bridges between two worlds; users don’t know when to use them instead of traditional methods. The main bottleneck is the lack of context – AI doesn’t understand users well enough (e.g., it doesn’t know the content of internal communication or work habits), so it can’t truly assist them effectively. The goal is for AI to have more context than users and provide timely suggestions when making decisions, essentially acting as an “superbrain” rather than replacing humans.
3. Computing power has become the “oil” of AI: More important than algorithms is money, energy, and supply chains
Altman stated that building large-scale computing capabilities is becoming one of the most expensive infrastructure projects in history. This means that AI competition is no longer just about algorithms; it also involves developing chips, building chip factories, ensuring power supply, and managing global supply chains. Additionally, financing these projects is a challenge. AI companies need to understand capital management, energy efficiency, and supply chain logistics, just as oil companies need to find oil and build transportation infrastructure.
4. OpenAI wants to be a platform, not a “versatile player”
OpenAI aims to be a platform that allows others to build applications on top of its technology. They have abandoned projects like Sora (a video generation tool) and Atlas (a browser) because they consume too much computing power and talent. The ideal model is a unified AI interface (like ChatGPT) with open APIs, enabling developers to create various applications (e.g., in e-commerce or healthcare). This approach allows OpenAI to benefit more people and avoid direct competition with its customers.
5. The dilemma of AI security: Fear of loss of control and concentration of power
Altman views security as a balance between two risks: AI becoming too powerful and losing control, and power being concentrated in the hands of a few companies/models. The solution is to iterate in the real world, release products, and collect user feedback, similar to how airlines track and improve from accidents. He believes that “safe products cannot be created in isolation.”
Final Thoughts
What is most impressive about Altman’s interview is his honesty: he admits that he hasn’t fully adapted to AI, and that its implementation is slower than expected. This is good news for ordinary people; there’s no need to worry about being replaced by AI tomorrow. Instead, we can gradually hand over repetitive and inefficient tasks to AI, freeing up time for critical thinking, creativity, and human interactions. After all, no matter how advanced AI becomes, people will still care about each other and prefer real experiences. More important than wondering when AI will change the world is whether you have already started using it to help with small tasks, such as drafting emails or organizing to-do lists. Change always starts with the details.