All Good Things Come to an End
Moore’s Law is an empirical relationship and prediction made by Gordon Moore in 1965. He posited that the number of components on an integrated circuit would double every year. By 1975, he updated this theory stating that the number would double every two years. This relationship has proven true and has becoming integral to the semiconductor industry and was a key component in the economic growth of the 1990’s and early 2000’s. It’s no wonder Moore ended up co-founding Intel in 1968.
The semiconductor industry has used this “law” to plan its R&D and push the envelope on developing better, faster technology. By doing this, the industry was able to make a few versatile chips that could be used in a broad set of contexts, allowing the companies to sell large quantities of these chips. Using these economies of scale and increasing demand, the semiconductor industry became very lucrative, which allowed it to then continue to invest in the next generation of chips. Almost all modern-day technology can be attributed to Moore’s law from high speed internet, smartphones, GPS, etc. By increasing computation power, techniques such as machine-learning can digest and analyze large data sets.
As we have somewhat recently entered a new decade (it still feels like 2020), it seems Moore’s Law may be on the brink of extinction. The development of better chips has now approached the constraints of quantum physics, making it prohibitively difficult and expensive to shrink the size of components of a chip. According to Technology Review, the cost of a fabrication plan is increasing by 13% each year and the number of companies involved in this technology has drastically shrunk. Furthermore, there is an issue with controlling the heat these chips produce that has caused a moratorium in 2004 on increasing “clock rates,” which controls how fast microprocessors can execute instructions (Nature). Intel does not seem to be giving up on its founder’s “law,” but may have to start thinking of new ways to innovate.
The rise associated with Moore’s Law allowed for inefficient algorithms because there was no pressure, at the time, to develop the most efficient code. Now, as the computer power is hitting a ceiling, there will be new innovations in making code more efficient. As noted by Technology Review , Neil Thompson, an economist, worked with colleagues to switch a program from Python to C and they were able to run a calculation 47 times faster. They then further tailored the code to run a process that took .41 seconds using C while it took 7 hours using Python.
Efficient programming is the new era of innovation. I am currently learning how to code in Python and SQL and am hoping to continue to explore other languages (hopefully C!), but I am quite envious of anyone who already is a professional programmer as this becomes a baseline skillset in the workforce. As a former teacher, I also hope to see increasing investment in computer science courses starting earlier in a child’s education, so they are best prepared for this growing opportunity.
Although efficient programming will improve computing power, the semiconductor industry is looking for other ways to stay ahead of the game. Instead of creating general purpose chips, companies are now creating specialized chips, specifically for industries that have the money and resources to invest in this complex technology. Neil Thompson, the previously mentioned economist, is concerned that this will lead to a decline in improving computers for general purposes and, out of concern for the economy, argues that the government should invest in improving more versatile chips.
Although there is no obvious successor to Moore’s Law, there is hope that new innovations will find the next breakthrough in the semiconductor industry. As noted in Nature, Daniel Reed, a computer scientist, made an analogy that I found quite fitting and helpful when thinking about the future of technology. He said that airplanes have continued to improve and yet ‘a Boeing 787 doesn’t go any faster than a 707 did in the 1950s — but they are very different planes,’ suggesting that although the raw computing power may not increase, there will be other avenues to innovate and improve the technology.
