Potential Onset of Another ‘AI Winter’

Concerns are rising in the AI sector about the potential onset of another 'AI winter'. Recently, a Bloomberg columnist questioned, 'Is the AI winter finally upon us?' while the British newspaper The Telegraph declared, 'The next AI winter is coming.' An AI winter refers to a period when enthusiasm for AI diminishes and investment declines. Historically, the AI field has experienced several such winters.

The recent discourse on an AI winter has been triggered by investor concerns that AI technology may be overhyped. In a worst-case scenario, an AI winter could coincide with the bursting of an AI-inflated stock market bubble, affecting the broader economy. Although there have been previous cycles of AI hype, the current situation involves investments in the hundreds of billions, potentially leading to significant repercussions.

OpenAI CEO Sam Altman recently startled markets by stating that some AI startups are grossly overvalued. A study from MIT found that 95% of AI pilot projects fail. Previous AI winters were often caused by academic research highlighting the limitations of AI techniques or difficulties in real-world application.

The U.S. and allied governments heavily funded AI research during the early Cold War. Competing approaches included hard-coded logical rules and perceptrons, the latter being the precursor to today's neural networks. Perceptrons learned rules from data rather than starting with logic.

What precipitates an AI winter is definitive evidence that the hype cannot be met. The first AI winter began in 1966 and 1969 with studies showing AI's inability to surpass human intelligence. Today, similar studies question AI's ability to meet expectations.

The current AI boom is not reliant on public funding. Most funding is private, and AI is widely deployed. However, AI systems remain complex and expensive, with uncertain returns on investment. Whether AI can automate critical processes for companies and governments remains uncertain.