MUMBAI - In a video from Google’s 2018 I/O developer conference, the company’s Chief Executive Sundar Pichai demonstrates how its Duplex assistant will be able to book appointments for its users.¹ The personal AI assistant (PAI) calls a salon, and the receptionist is unaware she is talking to a PAI assistant, not a human. The audience laughs when the AI says a perfectly timed ‘mm-hmm’ instead of a more generic ‘yes.’

Sadly, this could go down in history as Google’s Xerox moment, as it shut down Duplex right around November 30, 2022, when OpenAI unleashed ChatGPT on the world. Like Xerox, which created much of the computing technology in the 1960s, but lacked the imagination to take full advantage of it, Google - despite being the frontrunner in AI by a significant lead - has missed out on multiple chances in the AI domain that have instead been captured by others: OpenAI released Dall-E and ChatGPT to the world and achieved great popularity, while Google’s Imagen and LaMDA failed to inspire the masses in the same way.

Within weeks of ChatGPT’s launch Sundar Pichai announced a code red alert for his firm.² Microsoft went a step further, with plans to invest over USD10 billion in OpenAI, ChatGPT’s parent company, and incorporate its chatbot into Word, Excel, PowerPoint, and its other Office applications.

Why did Sundar Pichai dangle the carrot of a PAI assistant with Duplex in 2018 but fail to follow up? There are three main reasons for this. First, Google viewed Duplex as an assistant to business, not one that could be bundled with its LaMDA and become a PAI for everyone. Second, when companies become big, their focus tends to shift from innovation as a pleasure to innovation as a money-making venture for themselves and their shareholders. The company focused its AI developments on creating new revenue streams immediately, rather than solving problems. Even now, it sees ChatGPT as a dan