#553. Google CEO on Whether Google Is Falling Behind in AI

#553. Google CEO on Whether Google Is Falling Behind in AI

📝 Episode Summary

In this episode, we cloned the tech podcast *Hard Fork* **Sundar Pichai on Whether Google Is Falling Behind in A.I.**

Our guest is Google CEO Sundar Pichai. After Google I/O, he had a direct conversation with Kevin Roose and Casey Newton about the forefront of AI: Is Google falling behind in the AI race? Where does Gemini still lag in coding and Agent capabilities? Will traditional search be replaced by AI mode? How will AI affect graduates, doctors, programmers, and ordinary people's jobs? How should governments regulate AI? And how should companies like Google take responsibility as AGI and the "Singularity" draw closer?

In this episode, Sundar shows both confidence in Google’s future in AI and candidly acknowledges gaps in coding, long-cycle tasks, and Agent capabilities. He emphasized that coding will eventually become the foundation of all work; search will continue to evolve, but sources and links remain important; public anxiety about AI is completely reasonable because humans haven’t evolved to handle such drastic changes. On the topic of AGI, he also gave a rare clear judgment: the pace of progress over the past one or two years makes AGI seem more likely to be near rather than far.

👨‍💼 Guest

Sundar Pichai, CEO of Google and Alphabet. Since 2015, he has served as CEO of Google and later became CEO of Alphabet. He has long driven Google’s transformation toward an AI-first strategy, leading strategic initiatives in key business areas and technologies including search, cloud computing, Android, Chrome, TPUs, Gemini, and Google DeepMind.

⏱️ Timestamps

00:00 Intro & Podcast Overview

Where Google Stands in the AI Race

01:27 From Bard to Gemini: Is Google Still Catching Up?

04:09 Coding as the Main Battlefield: Why It Will Be the Foundation of All Work

04:56 Feedback on Gemini 2.5 Flash Release: Quality, Pricing, and Usage Limitations

06:11 Can Big Companies Stay Focused: How Google Advances Multiple AI Frontiers Simultaneously

AI in Search

07:11 The Biggest Change in 25 Years: Will Google Search Be Replaced by AI Mode?

08:47 Business Model in the Age of AI Search: Ads, Subscriptions, and User Value

Public Anxiety and the Future of Work

09:41 AI Backlash: Why Only a Few Think AI Is Overall Beneficial

11:52 Facing Graduates: How the Next Generation Will Deal with AI Impact

13:12 Will Jobs Disappear: How AI Changes Programmers, Doctors, and Radiologists

Realizing Agents

15:51 Everyday Agents: Spark and Personal Workflows

16:14 How Sundar Uses Agents for Meetings and Calendar Management

18:09 Trust Issues with Agents: Transparency, Control, and Safety Boundaries

Regulation, Safety, and Recursive Self-Improvement

18:35 How Governments Should Regulate AI: Balancing Innovation and Oversight

19:56 Recursive Self-Improvement: Can AI Safely Improve Itself?

21:05 What to Do When Approaching AGI: Why We Must Avoid Lab Competition States

Compute, TPUs, and Platform Strategy

21:35 Why Google Still Sells TPUs to Competitors

22:20 Cloud Business, Platform Ecosystems, and Next-Gen AI Hardware

AGI and the Singularity

23:25 How Sundar Views the Term AGI

24:29 Is AGI Closer or Further Away: Progress Over the Past Two Years Changed My View

25:11 At the Foot of the Singularity Mountain: How Google Understands AI’s Future Before 2030

26:16 Closing: Sundar Pichai Says Goodbye to Hard Fork

🌟 Highlights

💡 Google Admits: Coding Is a Frontier to Catch Up With

Sundar was candid, saying Google is strong in text, multimodal, speech, reasoning, and overall intelligence, but still lags in Agent-based coding, instruction following, and long-cycle tasks. He believes coding isn’t just a feature—it will become the foundational capability of all future work.

“I think coding will ultimately become the foundation of everything we do.”

🔍 Search Won’t Be Replaced Overnight by AI Mode

When asked whether the ten blue links would disappear, Sundar didn’t give an aggressive answer. He emphasized that Google will evolve the search experience gradually with users, not leap ahead too far. AI mode will become increasingly important, but sources, links, and connections to web content will still be part of search.

“I try not to run too far ahead of users.”

😟 AI Anxiety Is Reasonable

When the host mentioned that only about 16% of people think AI is overall good, Sundar said he fully understands public concerns. He believes AI is one of the most profound technologies humans will participate in shaping, and change is happening too fast—humans haven’t evolved to handle such drastic and intense shifts.

“Humans weren’t evolved to handle so much change.”

🧑‍⚕️ Work Won’t Just Disappear, It Will Be Redefined

Sundar doesn’t deny that AI will bring disruption, but he opposes overly pessimistic determinism. Using spreadsheets, coding, doctors, and radiologists as examples, he explained that AI will change many people’s starting points, giving more people access to abilities previously unavailable, and freeing professionals from low-value tasks.

“There will be many, many more people who can write code.”

🤖 The Key to Agents Isn’t Showmanship, But Trust

Sundar shared his own example of using agents: having it check his calendar and color-code events by private, health, and work categories. He thinks this already feels like science fiction, but the real breakthrough for agents lies in making users feel safe, transparent, and in control. Otherwise, if something goes wrong, users will retreat.

“You have to make users feel that way. Like when we first wanted someone to sit in the back seat of an autonomous car.”

🛡️ When Approaching AGI, Companies Mustn’t Fall Into Competition Mode

Talking about recursive self-improvement and AGI safety, Sundar said that if responsible labs approach such moments, they shouldn’t discuss internally alone but must involve broader societal forces. He especially emphasized that during critical stages of AGI, the industry must avoid competition states.

“In those stages of AGI, we must avoid competition states.”

⚙️ Why Google Still Opens TPU to Competitors

When asked why it doesn’t keep compute power for itself, Sundar explained that Google Cloud is inherently a platform business. Providing TPUs externally brings revenue and economies of scale, and feedback from external researchers helps improve next-gen hardware.

“When you’re running a platform, there’s a platform side to your business.”

⛰️ AGI May Be Closer, Not Farther

Sundar believes AI is inevitably making foundational progress toward AGI. While he won’t use radical language like some entrepreneurs, the pace of progress over the past year or two leads him to believe AGI is more likely to be near than distant. He also explained what Demis Hassabis meant by “at the foot of the Singularity Mountain”—in Google’s context, the Singularity refers to the arrival of AGI.

“The pace of progress over the past one or two years makes me think it’s more likely to be near, not far.”

🌐 Additional Podcast Info

This podcast was produced using original voice tones, which may sound odd in some places.

AI translation was used, so there may be some awkward phrasings;

If you’d like to hear other foreign-language podcasts in Chinese, feel free to contact WeChat: iEvenight