Human-Level AI Is Right Around the Corner—or Hundreds of Years Away

IEEE Spectrum gathers expert views on when machines could match human-level intelligence and how that might affect work, medicine, robotics, and society. The answers range from a few decades to well over a century, and some contributors question whether the target itself is clear enough to predict.

Reading notes

#

• Robin Hanson says brain-level computers could arrive within a century if brain emulation works, or in two to four centuries if it does not, and that they would replace humans in most job tasks.
• Martine Rothblatt says computers capable of the brain will arrive in the 21st century and argues that computerized versions of people could keep them effectively alive.
• Ruchir Puri says human brain traits such as caring, empathy, sharing, ingenuity, and innovation may remain elusive to machines for a long time.
• Ray Kurzweil predicts that computers will match and then quickly exceed human capabilities in areas where humans are still superior today by 2029.
• Kurzweil says AI is already helping with disease cures, crop yields, poverty reduction, and environmental problems, and that merging with AI is the best way to keep it safe and beneficial.
• Carver Mead argues that past technology fears have not ended society and that technology has spread prosperity and good overall.
• Nick Bostrom says human-level AI within a few decades seems credible, but worries about scalable control so advanced AI keeps behaving as intended as intelligence increases.
• Rodney Brooks says human-like intelligent and conscious machines are more than 100 years away, maybe many hundred years, and sees little point in trying to predict their effects now.
• Gary Marcus says machines already surpass brains in arithmetic and memory, but may still need 20 to 50 years to read, comprehend, and reason about novel situations like people do.
• Marcus says the larger worry is machines having too much power, and he expects progress in science and medicine if machines reach human-level reading and reasoning.
• Jürgen Schmidhuber says computing keeps getting cheaper and that cheap devices with human-brain raw computational power will come soon, with later systems spreading through space and beyond the solar system.