Google's leadership is gearing up for what they see as a pivotal year in artificial intelligence. At a recent internal strategy meeting, CEO Sundar Pichai didn't mince words about the company's priorities for 2025.
According to an audio recording obtained by CNBC, Pichai told employees that 2025 would be a defining year for the company. He stressed the importance of staying focused on using AI technology to solve real user problems.
"I think it’s really important we internalize the urgency of this moment, and need to move faster as a company. The stakes are high," Pichai said.
Pichai was candid about Google's position in the AI race. "In history, you don’t always need to be first but you have to execute well and really be the best in class as a product. I think that’s what 2025 is all about," he told employees.
During the meeting, DeepMind co-founder Demis Hassabis shared his vision for a universal AI assistant that would work "seamlessly operate over any domain, any modality or any device." The company plans to update an early version of this assistant, called Project Astra, in the first half of 2025. For now, Astra remains limited to select testers.
OpenAI may release an AI agent to control computers called "Operator" in January. Anthropic released a similar experimental product called Claude Computer Use in October.
No Plans for Test-Time Premium Subscription
When asked about potentially offering a $200 subscription similar to ChatGPT Pro, Hassabis was clear: "Right now, we don’t have any plans for this kind of subscription level."
He pointed out that their current $20 monthly Gemini Advanced subscription already provides good value, though he didn't completely rule out premium options in the future. "I wouldn’t necessarily say never."
This stance might offer insights into Google's AI strategy. While OpenAI's ChatGPT Pro offers an improved o1 model with additional test-time compute resources for scientific and professional applications - and plans to expand this with its upcoming o3 model - Google might either see limited potential in this approach or may not yet have comparable test-time compute capabilities.
In late December, Google unveiled several new AI models: Gemini 2.0 Flash, Flash Thinking, and the video generator Veo 2. According to Google's benchmarks, Veo 2 outperforms OpenAI's much-discussed Sora Turbo.
