Microsoft CEO Satya Nadella questions the value of AI
While Microsoft is investing tens of billions of dollars on data centres to support AI development, its CEO questions its value
Even Microsoft CEO acknowledges that despite AI improvements, the technology is still not delivering for business and consumers
Microsoft CEO Satya Nadella has questioned the value of artificial intelligence (AI), despite investing tens of billions of Microsoft’s dollars in the technology in various ways, suggesting its value will be demonstrated when it manifests in economic growth and wealth in people’s pockets.
In an interview with Dwarkesh Patel, an influential podcaster of which Nadella is a fan, the Microsoft CEO suggested that AI hadn’t achieved much, yet, and would only prove its value when solid efficiencies and economic growth could be attributed to it.
“The real benchmark: is the world growing at 10 per cent," he asked. "Suddenly productivity goes up and the economy is growing at a faster rate. When that happens, we'll be fine as an industry.”
What will count, he says, is industrial revolution-style irreversible growth.
However, most AI activity taking place in data centres today is focused on training rather than generative AI, and the AI applications currently running require ongoing supervision.
While AI is improving fast, it remains wildly flawed. For example, pioneer OpenAI recently admitted that its most advanced AI models are still not good enough compared to human developers. OpenAI founder Sam Altman nevertheless believes that models coming before the end of the year will be able to outperform low-level software coders.
Indeed, AI remains only as good as the data on which it was trained, and the increasing volume of data on the internet being generated by AI risks making it less, not more, reliable. Research indicates that AI models can collapse when trained on recursively generated data. And there have been suggestions that leading AI models have been cheating widely used benchmark tests.
Moreover, a recent Eye on the Market [PDF] report by Michael Cembalest, chairman of Market and Investment Strategy for JP Morgan Asset Management, questions whether the immense investments in AI and the infrastructure required to support it, already made or committed by the tech giants, will ever pay off.
For example, OpenAI is forecast to rack up $44 billion in financial losses before it reaches profitability in 2029. The CEO of Anthropic, the creator of the Claude AI, has suggested that by 2027 it will cost $100 billion to train just one AI model, such is the vast quantity of data and compute power required to create a reliable and workable model.
That makes open-source AI highly unlikely to emerge. There have also been warnings that a genuinely intelligent AI would be unpredictable, while the level of data available to train models is already running out.