But more engaging research can only be useful for Google when Google produces its most important output: profit. Typically, most customers are not willing to pay for AI features directly yet, so the company may want to sell ads in the Gemini app. Of course, this is Google’s classic strategy that spread to the rest of Silicon Valley a long time ago: Give us your data, your time and your attention, check out the box on our terms of service to save us from liability, and we won’t charge a dime for this cool tool you built.
Currently, OpenAI estimates that it has installed 600 million global applications for 140 million Google’s Gemini App, according to Sensor Tower. And, there are many other chatbots in this AI race – Claude, co-pilot, Grok, Deepseek, Llame, Cllexity, many of which are the largest and most funded competitors (or, for Claude, Google itself). The entire industry, not only Google, requires billions of dollars in investments for generated AI systems, so far without convening and a large amount of energy, enough to extend the lives of decades-old coal-fired power plants and nuclear reactors. The company insists that efficiency adds up every day. They also want to reduce the error to the point of winning more users. But no one really figures out how to produce reliable returns or avoid climate.
One challenge for Google is that its competitors do not: up to a quarter of search ad revenues may be lost due to antitrust judgments in the coming years. According to JP Morgan analyst Doug Anmuth. The priority of resubsidies will not lose anyone in the company. Some of Hsiao’s Gemini employees have been on vacation throughout the winter for three consecutive years to keep pace. Google co-founder Brin reportedly told some employees last month that they worked 60 hours a week, which was the “best option” to win the productivity of strengthening AI competitions. Among current and former employees who spoke with the link, there are concerns about more layoffs, more burnout and more legal troubles.
A Google researcher and a senior colleague said the general feeling was uneasy. The generated AI is obviously helpful. Even governments that are easy to manage large technologies such as France are preparing for the lofty commitment of the technology. In Google Deepmind and public negotiations, Hassabis did not relax an inch from his goal of creating artificial universal intelligence, a system capable of having human-level cognition in a range of tasks. He spent the occasional weekend and walked around London with his Astra prototype, savoring a future, from the Thames ducks to the entire physical world there to search for it. But AGI will require the system to get better in terms of reasoning, planning and responsibility.
In January, Openai took a step forward by letting the public perform another experiment: its long-standing carrier service, a so-called proxy AI that can go far beyond the chatbot window. An operator can click and enter a website on a website like a person, such as performing chores, such as booking a trip or filling out a form. Currently, it is much slower than what humans perform and with its unreliability (as part of a $200 per month plan) at a huge cost (available for). Naturally, Google is also working to bring agent functionality to its upcoming models. The current Gemini can help you with meal plans, and the next one will put your ingredients in your online shopping cart. Maybe the one after that will give you real-time feedback on onion switching technology.
As always, moving quickly can mean hanging out frequently. In late January, ahead of the Super Bowl, Google released an ad in which Gemini was caught in a more ridiculous mistake than Bard’s telescope error: It estimates half or more of all the cheese consumed on Earth is Guda. As Gemini grows from sometimes suspicious fact machines to an intimate part of human life – life coach, Panoramic Assistant – Pichai says Google is moving with caution. However, he and other Google executives finally never want to be caught again. The game continues.
Let us know what you think of this article. Submit a letter to the editor mail@wired.com.
