Is there an equivalent to Moore’s Law for AI? I don’t have an answer but I’ve started asking the question. My thinking was spurred by the throwaway line in the most recent issue of Spectrum from MIT:
The hottest ticket on the MIT campus this fall is…an undergraduate class in the Department of Electrical Engineering and Computer Science (EECS) whose humble course catalog label, 6.036, belies its exponentially growing popularity. The subject, Introduction to Machine Learning, was first offered in 2013 and now attracts hundreds more students than can fit into a 500-seat lecture hall. In addition to enrolling droves of EECS students, 6.036 brings in registrants from nearly every discipline MIT offers, from architecture to management. The irresistible draw? A chance to get a jump on the most powerful driver of technology innovation since Moore’s Law.
For 50 years, the tech industry has been powered by Moore’s Law. In 1965, co-founder of Intel Gordon Moore observed that the number of components per integrated circuit had doubled every 12 months since the 1950’s. He predicted that it would continue to do so for the next decade.1
Moore’s Law is not a law of nature like Newton’s laws of gravity; rather it was a historical observation and aspirational benchmark. Nevertheless, Moore’s law–combined with Dennard scaling–has been an essential factor in driving the tech revolution.
Moore’s Law may or may not be slowing down, but I’m interested in other times where exponential increase has presaged technological shifts that encourage entrepreneurial activity. Two other historical examples come to mind: (1) Metcalfe’s Law and the (2) the impact of the web’s stunning growth on a young Jeff Bezos.
In 1994, Bezos was working on Wall Street when he saw a research report predicting that the web would grow 2300% per year. Within months, he quit his lucrative hedge fund job in New York to move west and found Amazon. When Success Magazine asked him in 1998 why he moved so quickly, he replied:
“That figure of 2,300 percent,” he says. “That’s huge. Nothing usually grows that fast outside a petri dish. With that kind of growth rate, a sense of urgency becomes your most valuable asset.”
We don’t currently have a good definition of what the AI equivalent for Moore’s law is. I can propose some possibilities but I suspect that most of them will be less useful than Mark Zuckerberg’s 2008 coinage of Zuckerberg’s Law of Information Sharing which may be true but seems less revealing than Metcalfe’s or Moore’s.
Regardless of the actual metric we use for AI, it would be useful to come up with something that tracks the growing breadth, depth, and power of machine intelligence as it changes every aspect of the tech industry in the decades to come.
1. In the 1970’s, Moore revised the doubling period to every 18-24 months.↩