20 Questions to Understand DeepSeek and the "Second Half of AI" It Unleashed

Thanks to all of your support, Crossroads has had a remarkable year: as a newcomer podcast just one year in, we've won two major awards on Xiaoyuzhou — "New Podcast of the Year" and "Trending Podcast of the Year (Frontier Tech)." In the year ahead, Crossroads will continue to innovate and break new ground, moving forward actively with all of you in the AI era. This week, Crossroads launches its inaugural "20 Questions" column: twenty carefully designed heuristic questions that will offer a layered, in-depth look at the story behind DeepSeek — from technical breakthroughs to business impact to future trends.

Why "20 Questions"? Because when facing a major technological breakthrough like DeepSeek, we need a structured framework for thinking. These twenty questions aren't simple knowledge explainers; they're a carefully curated path of inquiry: from technical principles to industry impact, from phenomenon to prediction. We hope this Q&A format helps listeners build a clear, systematic understanding of DeepSeek amid today's information overload. Especially now, as the DeepSeek buzz has cooled somewhat, our discussion can be more rational and comprehensive — less swayed by short-term emotions.

This week, we're joined by Leding Li, who spent ten years as a Principal Senior Architect at Baidu and has previously appeared on Crossroads. His last episode, "I Want to Shatter Your Unrealistic Fantasies About AI, and Rebuild a Correct Understanding," became our most popular of 45 episodes last year. So we're honored to have Mr. Li back to discuss this topic as the DeepSeek fervor gradually subsides.

Our curated twenty DeepSeek-related questions span algorithms, compute, data, applications, and business. In a lifetime, major technological breakthroughs are few and far between — DeepSeek's release is undoubtedly one of them. I believe everyone owes it to themselves to invest the time to understand deeply: What exactly is DeepSeek? What does it mean? And how will it shape our future?

🟢 [20 Questions] Part 1: DeepSeek's Technical Breakthroughs and Innovation

02:26 Why does the release of DeepSeek-R1 mark AI's entry into the second half? What are the first and second halves of AI?

07:25 Explaining DeepSeek-R1 & DeepSeek-V3: How do they work, and how are they trained?

18:40 Has DeepSeek surpassed OpenAI? Do we have technology they don't? Did DeepSeek actually distill from OpenAI? Can distillation surpass SOTA?

21:43 DeepSeek-R1's RL vs. Anthropic's RLHF: a battle of approaches?

28:04 What factor mattered most in DeepSeek's victory — open source, technology, cost, or product?

42:45 How did DeepSeek-V3 cost only $5.5 million? And what's the story behind Fei-Fei Li training an AI reasoning model, s1, for $50 that rivals DeepSeek-R1 and OpenAI o1?

🟢 [20 Questions] Part 2: How to Correctly Understand DeepSeek's Impact?

51:39 How should we view DeepSeek's impact on China's tech industry, and Chinese business as a whole?

55:12 Why could Liang Wenfeng and DeepSeek pull this off? Why couldn't other major domestic companies?

56:54 DeepSeek's victory looks like a technology victory. What should PMs do in the AI era?

🟢 [20 Questions] Part 3: DeepSeek's Impact on Various AI Businesses and Applications

01:01:42 Impact on toC AI businesses?

01:04:24 Impact on toB (Enterprise & Government) AI businesses?

01:07:05 Impact on toDeveloper AI businesses?

01:09:15 Will it trigger a massive explosion of AI applications?

01:13:35 Will DeepSeek build its own chatbot?

🟢 [20 Questions] Part 4: "The Post-DeepSeek Era": What Changes Lie Ahead? What Will Remain Important? What Will No Longer Matter?

01:16:25 Will compute still matter in the future? Will giants' sustained investment change? What will happen to NVIDIA's stock?

01:18:26 Will data still matter in the future? Is it still a moat?

01:19:35 Predictions for DeepSeek's next moves?

01:22:01 Predictions for the next moves of China's major players? ByteDance, Tencent, Alibaba

01:23:33 Predictions for the next moves of U.S. major players? Meta, Google, Microsoft, Amazon, OpenAI, Anthropic

01:26:56 Using Peter Thiel's innovation theory to analyze DeepSeek/R1's success in low-cost inference: at what specific levels was synergistic innovation achieved?

Subscribe to the "Crossroads" podcast 🚦 We track the industry transformations and new entrepreneurial opportunities brought by the new wave of AI technology. Crossroads is Steve Jobs' metaphor for Apple — standing at the intersection of technology and liberal arts, where great products are born. AI is transforming industries across the board. We seek out, interview, and bring together "active agents" of the AI era, exploring and embracing new changes and new possibilities alongside them.

👦🏻 Host Koji: Co-founder of The Fair and Tangdao. I believe technology, especially AI, will fundamentally transform society and empower humanity in the future. Welcome to reach out, exchange ideas, and connect on what's next. Koji's Jike, Koji's website

👧🏻 Host Ronghui: Works at a tech VC, former Silicon Valley correspondent for Yicai Weekly. Ronghui's Jike

Join the "Crossroads" membership group ☀️ First-hand AI news and insights

👫🏻 We encourage you to date / make friends / find future kindred spirits

🦀 Add our assistant on WeChat to join: Rwkfbcianvd, or scan the QR code below

🎄 This podcast is supported by The Fair's Sound Forest Podcast Initiative