Summary of Key Developments
Google has recently experienced two significant changes in its AI leadership: First, Demis Hassabis, the founder of DeepMind, stepped back from frontline roles to become Chairman and Chief Scientist of the company, with Koray Kavukcuoglu, who has a more engineering-oriented background, taking over day-to-day operations. Second, four of Google's founding AI pioneers, including Jeff Dean, have left to establish a non-profit AI company called Discovery Loop. These developments come amid delays in the release of Google's Gemini 3.5 Pro model and increasing competition in the AI field, reflecting Google's transition challenges in the era of generative AI—namely, the shift from long-term scientific exploration to short-term product development, as well as the tension between large corporate structures and the vision of top researchers.
1. The Delay of Gemini: A Major Setback for Google's AI Competitiveness
Gemini was once Google's trump card against ChatGPT, but recent setbacks have dampened its momentum. When it was released at the end of 2023, Gemini Ultra was touted as surpassing GPT-4, and the 2.5 Pro version did show promise. However, the much-anticipated 3.5 Pro has yet to appear, despite being scheduled for June. This is critical for Google, as generative AI competition focuses on who can release new models faster and generate revenue more quickly, rather than just publishing research papers in labs.
Why the delay? It's not that Google lacks the technology; rather, there are internal issues such as poor coordination, inefficient resource allocation, and a disconnect between research and product development. By appointing Koray Kavukcuoglu, who has experience managing chip development, recruitment, and team restructuring, Google is signaling a shift towards more pragmatic approaches. In other words, they are prioritizing the practical implementation of models over theoretical work.
2. The Transformation of DeepMind: From Scientific Ambition to Commercial Reality
DeepMind was once hailed as the pinnacle of Google's AI efforts, thanks to groundbreaking projects like AlphaGo (which defeated human Go players) and AlphaFold (which solved protein structures). Its founder, Demis Hassabis, focused on solving complex intellectual problems rather than immediate profit generation. However, things have changed:
- For Gemini, Google disbanded the AlphaFold team, reassigning its researchers to work on Gemini or AI-related pharmaceutical research.
- Hassabis has moved from CEO to Chairman, indicating a shift in DeepMind's focus from long-term scientific breakthroughs to shorter-term product iterations.
In simple terms, DeepMind, once a hub for scientific innovation, is now being transformed into a factory that produces AI products for commercial use.
3. The Departure of Key Researchers: A Matter of Different Priorities
The departure of Jeff Dean and his colleagues is not a sign of conflict but rather a reflection of diverging goals. These individuals played key roles in developing Google's big data processing technologies. They left to found Discovery Loop, a non-profit dedicated to using AI to address critical scientific and engineering challenges (such as climate change and disease prevention), with the aim of making a positive impact rather than maximizing profits.
The reason for their departure may be the discrepancy between Google's commercial focus and their own aspirations. While Google wants to quickly monetize Gemini, these researchers prefer more pure scientific exploration. Nevertheless, Google has continued to invest in Discovery Loop, indicating a mutually beneficial arrangement where Google can benefit from their innovative work while they pursue their vision of using AI for social good.
4. The Dilemmas Faced by Google in AI
Google possesses some of the best resources in the world—computing power, data, and funding—but these advantages also pose challenges:
- Slow decision-making due to the need to coordinate across multiple departments (search, cloud, DeepMind).
- Short-term pressure from shareholders to achieve profitable results.
- A clash of visions between top researchers aiming for groundbreaking projects and the company's focus on quick profit generation.
Smaller teams like Discovery Loop are more flexible, with fewer resources and clear goals, allowing them to focus on their research without being constrained by corporate priorities. This is a common pattern in the AI industry, where large companies handle product development, while smaller teams explore new technological possibilities.
What Lies Ahead for Google's AI?
In the short term, the appointment of Koray Kavukcuoglu could accelerate Gemini's progress. In the long run, the loss of key researchers may impact Google's innovation capabilities. However, Google has made preparations by investing in Discovery Loop to maintain connections with these talented individuals.
These changes represent a fundamental choice for the AI industry: large companies must balance profitability with technological advancement, while top researchers strive to pursue their scientific ambitions. Whether Google or its departing colleagues will emerge victorious remains to be seen. But one thing is certain: both parties can benefit—Google through its products, and the new companies through their innovative breakthroughs, which could ultimately lead to better AI solutions for users.