Google’s response to OpenAI’s ChatGPT launch sparked a dramatic two-year transformation within the tech giant, according to an extensive report by Paresh Dave and Arielle Pardes in Wired. The article details how Google, despite having developed transformer-based architecture (the “T” in ChatGPT), found itself playing catch-up in the generative AI race.
When ChatGPT debuted in December 2022, Google declared a “code red” and tasked executive Sissie Hsiao with building a rival product in just 100 days. This effort led to Bard, which initially stumbled when it made a factual error about the James Webb Space Telescope during its public unveiling, causing Alphabet’s shares to drop by 9 percent.
To regain its competitive edge, Google merged its two AI research teams – DeepMind and Google Brain – into a single entity led by Demis Hassabis. This consolidated team developed Gemini, a more powerful language model designed to outperform ChatGPT. According to James Manyika, Google’s SVP of technology and society, the company made “bold and responsible choices” throughout this process.
The race to deploy AI products quickly led to significant cultural shifts within Google. The company conducted its first mass layoffs in history, cutting 12,000 jobs. Teams working on AI products faced immense pressure, with many employees working nights and weekends. Google’s typically thorough safety review processes were condensed to accommodate accelerated launch timelines.
“It’s been like sprinting a marathon,” Hsiao told Wired. The company’s aggressive AI push appears to have paid dividends for investors, with Alphabet shares nearly doubling from their low point after ChatGPT’s debut.
The article highlights how Google grappled with various challenges, including hallucinations and inappropriate responses in its AI models. When Gemini’s image generation feature launched in February 2024, it faced criticism for producing historically inaccurate depictions, such as showing diverse individuals in roles traditionally held by white men, like U.S. senators from the 1800s.
Google has since integrated Gemini capabilities across its products, including Search, Maps, and YouTube. The company introduced AI Overviews in Search, which synthesizes search results but has occasionally produced bizarre recommendations, such as suggesting adding “non-toxic glue” to pizza sauce.
Despite these setbacks, Google executives express optimism about the company’s AI future. Hassabis believes Google has “the broadest and deepest research base” of any organization and is pursuing ambitious goals like curing diseases with AI. CEO Sundar Pichai maintains that Google is proceeding cautiously as Gemini evolves from a facts machine to a more intimate part of users’ lives.
The competitive landscape remains challenging, with OpenAI’s ChatGPT boasting approximately 600 million app installations compared to Gemini’s 140 million, according to data from Sensor Tower. Additionally, Google faces potential revenue losses from antitrust judgments, increasing pressure to monetize its AI investments.