The Race Google Almost Won
Mustafa Suleyman, a co-founder of the AI lab DeepMind which Google acquired, and now the CEO of Microsoft AI, has pulled back the curtain on a pivotal moment in tech history. In his book, “The Coming Wave,” and in recent interviews, he describes a period
of intense frustration within Google before the public had ever heard of generative AI. According to Suleyman, Google possessed technology in its LaMDA (Language Model for Dialogue Applications) chatbot that was essentially “ChatGPT before ChatGPT.” Internally, a debate raged. One faction of engineers and executives, including Suleyman himself, was “desperate” to release the groundbreaking technology to the public, believing it represented the future of search and human-computer interaction. They saw the immense potential and were eager to be the first to introduce it to the world. However, they were met with powerful resistance from another camp within the company.
A Culture of Caution
The resistance to launching LaMDA wasn't born from a lack of vision, but from a deep-seated sense of caution. Google, as the established king of information with a multi-trillion-dollar valuation, had far more to lose than a startup like OpenAI. The potential for reputational damage from an AI that produced biased, toxic, or simply incorrect information was enormous. Internal testing of early versions of Google's chatbots revealed significant flaws, with one employee reportedly calling a pre-release version of Bard a “pathological liar.” These concerns were not new. Years prior, Google had been rocked by internal controversies, most notably the ousting of its ethical AI co-leads, Timnit Gebru and Margaret Mitchell, after they co-authored a paper warning about the dangers of large language models, the very technology powering LaMDA and ChatGPT. This history created a corporate culture where the brakes were often applied to projects deemed too risky, and Suleyman notes the other half of the company just “couldn’t get their heads around” releasing it.
The 'Sentient' AI Sideshow
The charged atmosphere inside Google was thrown into the public spotlight in mid-2022. A Google engineer named Blake Lemoine went public with the extraordinary claim that he believed LaMDA was sentient, possessing the self-awareness of a human child. He published conversations he had with the chatbot that he felt proved its consciousness. While the broader scientific community largely dismissed the claim, the incident highlighted the profound and often unsettling power of these new models. It demonstrated just how convincingly a large language model could mimic human conversation, raising complex ethical questions about the nature of intelligence and consciousness. For Google, it was another public relations headache that reinforced the arguments for a go-slow approach, even as competitors were preparing to leap ahead.
The Cost of Hesitation
While Google debated, OpenAI acted. The launch of ChatGPT in November 2022 completely blindsided the public and reset the narrative in the tech industry. Overnight, OpenAI, backed by Microsoft, was seen as the leader in a new technological revolution. Google, despite its years of pioneering research, was suddenly on the back foot. The company scrambled to release its own version, Bard, but the rushed debut was marred by a factual error in a promotional video that wiped a reported $100 billion off the company's market value. Suleyman’s account confirms that Google’s hesitation, rooted in a responsible if perhaps overly cautious ethical framework, cost it the coveted first-mover advantage. The company that had the technology first ended up looking like a follower, a perception it has been fighting to change ever since.
