When will computer hardware match the human brain? (1998) (www.jetpress.org)

In a 1998 study, Hans Moravec proposed that computer hardware might match the human brain's intellectual capabilities by the 2020s, hinging his argument on the correlation between AI performance and hardware advancements. The paper estimates that approximately 100 million MIPS (Million Instructions Per Second) would be needed to replicate human behavior, drawing on specific calculations related to processes in the retina and comparing them to the capabilities of existing AI technologies in fields like computer vision and speech recognition. The research underscores the need for increased processing power to enable more sophisticated learning mechanisms in AI systems. This prediction is significant for the AI/ML community as it highlights the crucial interplay between hardware evolution and AI development, implying that as processing power becomes more accessible and affordable, the sophistication of AI will similarly grow. Moravec outlined various benchmarks, suggesting that for complex tasks, higher MIPS would facilitate learning-driven improvements in AI performance, ultimately leading to machines that could independently handle tasks akin to human cognitive functions. The implications of these findings remain relevant today, prompting ongoing discussion about the future of AI capabilities in light of an ever-increasing demand for computational efficiency.