Are chatbots just passing guests? Google bets on "world models," hoping smart glasses become the true "killer" application of AI

Google is adjusting its strategic focus in the field of artificial intelligence, attempting to move beyond the currently dominant chatbot paradigm by betting on "world models" that can understand the physical world, seeking the next qualitative leap in AI technology.

According to Bloomberg columnist Parmy Olson, Google plans to launch a new AI smart glasses next year (2026), aiming to compete with differentiated competition against rivals like Meta through "world model" technology. These glasses will be manufactured in collaboration with Samsung, and unlike competing products that can only describe scenes through cameras, this device aims to understand three-dimensional space, the relationships between physical objects, and environmental dynamics. This is seen as a key attempt by Hassabis in the product field, expected to reverse Google's past reputation in the smart glasses sector and establish new industry standards.

This move comes as Google gradually regains ground in the AI race. With the successful release of the Gemini 3 model, Google has topped the performance rankings and made strong strides in user scale, effectively catching up to OpenAI, forcing competitor Sam Altman to respond to market pressure by releasing "red code." Although the chatbot ChatGPT still leads in enterprise business and first-mover advantage, Google, with its substantial resource reserves and dual-track technology approach, is redefining the landscape of this "tortoise and hare" race.

Market views suggest that if this world model-based smart glasses can achieve commercial success, it will not only signify a revival of Google's hardware business but may also mark a shift in AI applications from mere language processing to interactions with the physical world. For investors, this is not only about whether Google can find a "killer app" in the AI era but will also determine whether Hassabis can successfully transition from a top scientist who has won a Nobel Prize to a business architect defining the next era of Google.

Beyond Chatbots: Strategic Bet on "World Models"

In Google's strategic landscape, large language models (LLMs) represented by ChatGPT are not the only path to artificial general intelligence (AGI). Although OpenAI and Meta's Zuckerberg are fully betting on chatbots trained on vast amounts of web content, investing hundreds of billions in computing power, Hassabis has always insisted that "world models" trained on simulations and surrounding physical environments will lead to the next leap in AI.

This divergence in technological philosophy has begun to emerge in the industry. Hassabis advocates that AI should possess the ability to understand the physical world, such as the Project Astra initiative aimed at enabling AI to comprehend how objects move in space and their interrelationships.

In contrast, Meta's AI chief Yann LeCun, while holding similar views, left due to an inability to reach consensus with Zuckerberg. While competitors focus on pursuing super-intelligent chatbots, Google is diversifying its bets like a hedge fund, investing in both existing chatbot technology and heavily laying out plans for potentially paradigm-shifting technologies

The Comeback of Gemini and the Deep Integration of Organizational Structure

In order to stand out in fierce competition, Alphabet CEO Sundar Pichai merged the company's two major AI departments in 2023, placing them under the unified leadership of Hassabis. This move broke the long-standing opposition between the departments. Subsequently, in August 2024, Google effectively spent $2.7 billion to rehire Noam Shazeer, the co-inventor of the Transformer architecture. Shazeer's departure was due to Google's refusal to release his chatbot program, and his return proved to be crucial.

According to media reports citing informed sources in Silicon Valley, Shazeer was appointed as the technical co-lead of the Gemini chatbot upon his return and successfully identified and fixed a deep vulnerability in Gemini, greatly improving training efficiency. This directly contributed to the success of Gemini 3, which surpassed ChatGPT in performance benchmarks and boasted over 650 million monthly active application users, along with approximately 2 billion users acquired through Google Search.

Although Hassabis himself is not as obsessed with the LLM path as Shazeer, he skillfully resolved differences through diplomatic means, bringing Google back to the center of the race. "While a significant part of his personality is focused on science, an equally important part is focused on victory," commented Sebastian Mallaby, author of "The Infinity Machine," on Hassabis.

Searching for "Killer" Applications: The Commercialization Challenge of Smart Glasses

Despite Gemini's phased victory, Google still faces immense commercialization pressure. Currently, Gemini's enterprise version business was launched late, and scientific achievements like AlphaFold have yet to be transformed into FDA-approved drugs. Google urgently needs to prove that its AI technology can monetize through means other than advertising. The smart glasses planned for release next year carry this heavy expectation.

This device, developed in collaboration with Samsung, is expected to be equipped with lens displays for functions such as navigation and translation. More critically, if Hassabis can successfully apply the "world model" to this, the glasses will have the ability to remember the location of items (such as "where are the keys"), understand three-dimensional environments, and predict subsequent dynamics. This would create a generational gap in functionality compared to Meta's Ray-Ban glasses, which can only describe what is seen through the camera and lack deep processing capabilities of the physical environment.

If this new pair of glasses can operate successfully and gain market recognition, it will not only erase the poor reputation left by Google Glass but may also become a true "killer" application in the AI field, establishing Google's leadership in the next generation of computing platforms