BY RODNEY BROOKS
THE NEXT 50 YEARS
Expect new ways to understand computation, computational abstractions for our
computing machinery, and connections between people and their information
sources, as well as each other.
Since this issue celebrates 50 years of Communications it seems appropriate to speculate what
the next 50 years may bring, in time for the 100th anniversary. For this look into the future, I
cover three areas of “computing machinery”: the theoretical understanding of computation; the
technological substrate of computing machines; and the applications that computing machines
bring to our lives.

Our current understanding of computation sprang from Alan Turing’s famous 1936 paper on computable numbers but really got going around the time CACM was first published. It has centered around what is in principle computable and how much time and how much memory are needed to compute it.

By the early 1960s the idea of asymptotic analysis of algorithms was in place; if you can characterize some computation problem by a positive integer n (such as how many records you want to sort on some key) then what is the theoretical minimum time and/or storage space as a function of n required to complete the task? Even better, what is the algorithm for doing it? (In the case of sorting records, the time requirement is proportional to n*(log n), and the space is proportional to n.) This approach to understanding computation has been wonderfully productive and is still yielding both theoretical insights and practical results (such as in the field of cryptography). Other

AARON EDSINGER AND HIS DOMO UPPER-TORSO HUMANOID ROBOT ( 29 DEGREES OF FREEDOM) DEVELOPED AT THE MIT COMPUTER SCIENCE AND ARTIFICIAL INTELLIGENCE LABORATORY ( WWW.AI.MIT.EDU/PROJECTS/ HUMANOID-ROBOTICS-GROUP/).