Machine Intelligence: Come So Far, Got So Far To Go

Part 1 of "Separating Hype from Reality," Chapter 2 of The Machine Intelligence Primer

Recent advances in MI are awe-inspiring—particularly for the many researchers who’ve dedicated decades to seeing the technology to fruition and endured its many setbacks. New developments in MI research and enabling trends have made possible astounding technological feats: driverless cars, robotic surgeons, software that can compose music, write news articles, and translate languages in real-time. Amid a steady drumbeat of announcements over MI’s latest achievements and applications, a sense that humans may soon become passé has begun to take hold in certain circles. MI experts would say, “Not so fast.” 

Andrew Ng, one of the researchers most responsible for MI’s breakthroughs in deep learning, said recently, “The types [of MI] being deployed are still extremely limited...If a typical person can do a mental task with less than one second of thought, [MI] can probably automate it.” ¹ Stuart Russell, a professor at the University of California, Berkeley who has advanced MI research for decades, said this year, “I think we have at least half a dozen major breakthroughs to go before we get close to human-level [MI].” ²

In this section, we’ll give you an understanding of these researchers’ pragmatism by explaining what MI can and can’t do given the state of the technology. To start out, consider the following facts about what MI can and cannot do today—and notice which list is longer. (Fig. 2.1).