However, a growing body of intriguing research is only deemed beneficial to Google if it yields the ultimate result: profit. Generally speaking, most consumers are not yet inclined to pay for AI features directly, prompting the company to consider advertising within the Gemini app. This is a time-honored tactic for Google, which has long since pervaded Silicon Valley: Exchange your data, time, and attention for our services while agreeing to our terms of service that indemnify us from liability, and you won’t have to spend a cent on the innovative tool we’ve created.
Currently, as per Sensor Tower’s data, OpenAI’s estimated 600 million total global app installations for ChatGPT overshadow Google’s 140 million for the Gemini app. Numerous other chatbots are competing in this AI landscape, such as Claude, Copilot, Grok, DeepSeek, Llama, and Perplexity—many of which are supported by Google’s major and well-capitalized competitors (or, in the case of Claude, by Google itself). The industry as a whole, not just Google, grapples with the reality that generative AI systems have necessitated billions in investment, expenses that remain to be recouped, alongside immense energy consumption, enough to prolong the operational lives of aging coal plants and nuclear reactors. Corporations assert that efficiencies are accumulating daily. They also aspire to reduce errors sufficiently to attract a larger user base. Yet no one has definitively cracked the code for achieving a sustainable return or mitigating environmental impact.
Moreover, Google faces a challenge that its rivals do not: In the next few years, as much as a quarter of its search advertising revenue may be jeopardized by antitrust rulings, according to JP Morgan analyst Doug Anmuth. The urgent need to replenish finances is a reality acknowledged across the company. Some of Hsiao’s Gemini team members have worked through the winter holidays for three consecutive years to keep pace. Google cofounder Brin reportedly informed employees last month that 60 hours of work per week represents the optimal level of productivity necessary to prevail in an escalating AI competition. A sense of dread regarding further layoffs, increased burnout, and looming legal issues is pervasive among current and former employees who spoke to WIRED.
An internal researcher and a senior executive at Google express a dominant sentiment of unease. There’s no doubt that generative AI is advantageous. Even governments that typically regulate large tech companies, like France, are becoming more receptive to the astounding promises of the technology. Within Google DeepMind and in public forums, Hassabis remains unwavering in his commitment to achieving artificial general intelligence—an AI system with human-like cognitive capabilities across various tasks. Occasionally, he takes weekends to stroll through London with his Astra prototype, envisioning a future where everything in the physical world, from that duck in the Thames to this Georgian estate, is searchable. However, AGI will require advancements in reasoning, planning, and management.
In January, OpenAI made a significant stride toward that future by unveiling its highly anticipated Operator service—an agent-like AI capable of performing tasks far beyond the confines of a chatbot. Operator can navigate and interact with websites just like a human to accomplish tasks such as booking a trip or filling out forms. For now, it executes these actions considerably slower and more cautiously than a person, and its unreliability comes at a steep price (available under a $200 monthly subscription). Unsurprisingly, Google is also working to incorporate agent-like features in its forthcoming models. While the current Gemini can aid you in crafting a meal plan, the next iteration may automatically add your ingredients to an online shopping cart. Perhaps the version after that will provide real-time advice on your onion-chopping skills.
As always, haste can lead to errors. In late January, ahead of the Super Bowl, Google released an advertisement in which Gemini made an embarrassing mistake, even more comically erroneous than Bard’s telescope blunder: It stated that half or more of all cheese consumed globally is gouda. As Gemini evolves from a sometimes-reliable source of information to an integral aspect of human life—acting as a life coach and omnipresent assistant—Pichai asserts that Google is advancing with caution. However, having regained its leading position, he and other executives at Google might not want to risk being overtaken again. The competition continues.
Updated 3/21/2025, 4 PM EDT: Wired has clarified the context of a quote attributed to Pandu Nayak.
Share your thoughts on this article. Send a letter to the editor at mail@wired.com.
