Pursuing generative AI is a costly endeavor, primarily due to its high demand for computing power and the skilled talent required for advancements. Recent claims suggest that scaling laws may be plateauing, which could hinder the progress of leading AI labs such as OpenAI, Anthropic, and Google.
Microsoft and OpenAI have recently reaffirmed their multibillion-dollar partnership under a new definitive agreement. This arrangement enables Microsoft to pursue Artificial General Intelligence (AGI) either independently or in collaboration with other parties.
During a recent podcast episode of “Moonshots with Peter Diamandis,” Microsoft AI CEO Mustafa Suleyman mentioned that it could cost “hundreds of billions of dollars” to keep pace with AI development over the next five to ten years. He also pointed out the significant expenses associated with hiring individual researchers and technical staff.
Suleyman described Microsoft as akin to a “modern construction company,” highlighting the vast workforce required to build advanced computing infrastructure. He acknowledged that while developing frontier AI models is costly, large companies like Microsoft benefit from certain structural advantages.
Despite these challenges, Microsoft remains committed to AI. A leaked internal memo suggests that CEO Satya Nadella is directing the company towards a renewed focus on intelligence, integration, and AI. Suleyman noted that although Bill Gates’ past vision for a software factory was effective, it is no longer sufficient to meet current demands.
Nadella is reportedly urging top executives to accelerate their work on AI projects, emphasizing the need for efficiency. Insiders claim that Microsoft is struggling to balance innovative AI tools with the performance of its other products.
Following the new agreement with OpenAI, Suleyman asserted that Microsoft must become self-sufficient in AI. To achieve this, the company aims to train frontier models using its own data and state-of-the-art computing resources.
Microsoft has initiated a new MAI Superintelligence team to drive this effort. Suleyman expressed a clear ambition: “We’re absolutely pushing for the frontier. We want to build the best superintelligence and the safest superintelligence models in the world.”
This news arrives as investor enthusiasm for generative AI appears to be declining, with concerns about the substantial investments made in the sector without a defined revenue strategy.
As Microsoft embarks on this new phase of AI development, Suleyman has indicated that the company would reconsider its AI pursuits if they pose a significant risk to humanity, aligning with his vision for responsible superintelligence.
—
