The recent revelations stemming from the Department of Justice’s antitrust case against Google shed light on the intricate dynamics driving the AI landscape today. Unveiled emails expose a mixture of fear, envy, and fervent capitalist ambition brewing within tech giants. Among them, Microsoft emerges as a central player, grappling with Google’s formidable lead in artificial intelligence. The disclosed correspondence reveals Microsoft executives expressing alarm over Google’s AI advancements, catalyzing a sense of urgency within the company.
This urgency ultimately fueled Microsoft’s substantial investment in OpenAI, now recognized as a pivotal partnership. In a redacted 2019 email exchange titled “Thoughts on OpenAI,” Microsoft CEO Satya Nadella underscores the significance of the initiative, forwarding a compelling message from CTO Kevin Scott to CFO Amy Hood. Scott’s concerns about Google’s exponential AI growth, initially dismissed as mere “game-playing stunts,” underscore the competitive landscape’s high stakes. Google’s groundbreaking achievements, such as AlphaGo’s victory over world champion Ke Jie in 2017, served as a wake-up call, propelling industry giants like Microsoft to take decisive action in the race for AI supremacy.
Scott’s reflection on Microsoft’s attempts to emulate Google’s advancements in AI paints a vivid picture of the challenges faced in this competitive landscape. Initially dismissing Google’s progress as mere “game-playing stunts” proved to be a miscalculation, as Scott admits. The turning point came when Google harnessed its infrastructure to develop sophisticated natural language models, notably BERT-large, which eluded easy replication by Microsoft. Scott’s candid admission reveals the stark reality of the infrastructure gap between the two tech giants, highlighting the formidable lead Google had established in AI capabilities. Despite having the blueprint for BERT-large, Microsoft grappled with significant hurdles in model training due to infrastructure limitations, taking approximately six months to achieve what Google had already accomplished. This disparity underscores the critical role of infrastructure in AI development and serves as a cautionary tale for companies vying for dominance in this rapidly evolving field.
Scott’s admiration and envy of Google’s advancements extend beyond natural language models to encompass Gmail’s auto-complete capabilities, which he describes as “scarily good.” This sentiment underscores the significant strides Google has made in leveraging machine learning to enhance user experiences. In contrast, Scott acknowledges Microsoft’s lag behind the competition in terms of machine learning scale, emphasizing the need to bridge this gap to remain competitive. Despite boasting “very smart” individuals on its machine-learning teams, Microsoft faces challenges in scaling up its ambitions due to constraints within core deep learning teams. Scott’s candid assessment highlights the intricate balance between talent and infrastructure necessary for success in the AI arena. Moreover, his recognition of the growth trajectories of OpenAI, DeepMind, and Google Brain reflects the dynamic landscape of AI research and development, where innovation and competition drive progress. As Microsoft navigates these challenges, Scott’s insights serve as a roadmap for prioritizing investments and fostering a culture of innovation to propel the company forward in the AI race.
