In many areas – eg, effective computer speed, genomic sequencing, etc – there has been exponential increases in speed and/or exponential decreases in cost. The law of accelerating returns is a hypothesised generalisation in which over time more and more processes become automated processes and hence experience exponential increases in speed and/or exponential decreases in cost.
The law of accelerating returns is primarily espoused by Ray Kurzweil, although similar ideas have been proposed by others. It can be said to have two main elements:
Any automated process will experience continual exponential increases in speed and/or exponential decreases in cost.
More processes will become capable of automation (due to increases in the capability of technology in general and information technology in particular) and hence subject to 1.
Kurzweil presents evidence for this law holding historically, although some claim that both his data-point selection and “statistical” interpretation are flawed in ways that over-estimate the support for his assertions. He also provides scenarios for the law holding in the future, but again some dispute these.
A major plank of Kurzweil’s argument involves technological feedback effects: development of new technological capabilities can be used to do even more development at a quicker rate, resulting in an increasing (and, it’s argued, exponentially growing) rate of development. (An example would be the use of hand-designed computer chips to run computer-chip design software which is quicker and more accurate than the original hand design methods.)
Law of accelerating returns, Wikipedia.
TODO