Revisiting Kevin Kelly's 2016 AI Predictions From The Inevitable

In 2016, Kevin Kelly published The Inevitable, a book outlining macro technological trends and following them to their logical conclusions. A decade later, several of his AI related predictions stand out as remarkably prescient, while at least one major development escaped his foresight.

Google's AI Bet

Kelly recounts a 2002 conversation with Google cofounder Larry Page at a private party before the company's IPO, when Google was a small company focused only on search. Asked how free web search could sustain a business, Page replied, "Oh, we're really making an AI." Kelly notes that Google has since bought 13 AI and robotics companies in addition to DeepMind. Rather than using AI to make search better, Kelly argues, Google is using search to make its AI better. Each of the 3 billion queries Google conducts daily tutors the deep learning system, and actions like clicking the most Easter Bunny looking image teach it what things look like.

In a fall 2015 quarterly earnings call, CEO Sundar Pichai stated that AI was going to be "a core transformative way by which we are rethinking everything we are doing." Kelly's boldest prediction followed: by 2026, Google's main product will not be search but AI. The 2017 paper Attention Is All You Need, which introduced the transformer deep learning architecture, arrived a year after the book and underpinned much of what followed.

Cloud Intelligence and the Oligarchy

Kelly predicted that cloud based AI would obey the law of increasing returns, also called the network effect. The more people use an AI, the smarter it gets, and the smarter it gets, the more people use it. Once a company enters this virtuous cycle, it grows so big so fast that it overwhelms upstart competitors, leading to what Kelly called an "oligarchy of two or three large, general purpose cloud based commercial intelligences."

He also described the AI on the horizon as looking more like Amazon Web Services: cheap, reliable, industrial grade digital smartness running behind everything, almost invisible except when it blinks off. Ideally, Kelly argued, this additional intelligence should be not just cheap but free, like the free commons of the web. He advocated for open weights, a position that remains central to AI policy debates today.

Narrow Minds and Conversational Answers

Kelly conceded that he did not foresee the bitter lesson, the idea that general methods leveraging computation and data ultimately win. He predicted that the bulk of AI would be special purpose software brains: programs that translate any language into any other language but do little else, drive a car but cannot converse, or recall every pixel of every video on YouTube without anticipating work routines. In the next 10 years, he wrote, 99 percent of the artificial intelligence people interact with, directly or indirectly, would be narrow, supersmart specialists.

He also foresaw a world where anyone can ask the cloud, in conversational tones, any question at all, and receive an explanation if a known answer exists. Questions would range from trivia like who won the 1974 Rookie of the Year Award to everyday needs like where Jenny is or when the next bus arrives. With the manufacturing cost of each answer at nanocents, Kelly concluded, search would no longer be considered a first world luxury. It would become an essential universal commodity.

Sources: taylor.town