Given machine learning, as a subject, has existed for many decades, it begs the question: why hadn't it become as popular as it is today much sooner? Indeed, the theories of complex machine learning algorithms such as neural networks were well known by the late 1990s, and the foundation had been established well before that in the theoretical realm.
There are a few factors that can be attributed to the success of machine learning:
- The Internet: The web played a critical role in democratizing information and connecting people in an unprecedented way. It made the exchange of information simple in a way that could not have been achieved through the pre-existing methods of print media communication. Not only did the web transform and revolutionize the dissemination of information, it also opened up new opportunities. Google's PageRank, as mentioned earlier, was one of the first large-scale and highly visible successes in the application of statistical models to develop a highly successful web enterprise.
- Social media: While the web provided a platform for communication, it lacked a level of flexibility akin to how people interacted with one another in the real world. There was a noticeable, but understated, and arguably unexplored gap. Tools such as IRC and Usenet were the precursors to social network websites such as Myspace, which was one of the first web-based platforms intended to create personal networks. By early-mid 2000, Facebook had emerged as the leader in social networking. These platforms provided a unique opportunity to leverage the Internet to collect data at an individual level. Each user left a trail of messages, ripe for collection and analysis using Natural Language Processing and other techniques.
- Computing hardware: Hardware used for computers developed at an exponential rate. Machine learning algorithms are inherently compute and resource intensive, that is, they require powerful CPUs, fast disks, and high memory depending on the size of data. The invention of new ways to store data on solid state drives (SSDs) was a leap from the erstwhile process of storing on spinning hard drives. Faster access meant that data could be delivered to the CPU at a much faster rate and reduce the I/O bottleneck that has traditionally been a weak area in computing. Faster CPUs meant it was possible to perform hundreds and thousands of iterations demanded by machine learning algorithms in a timely manner. Finally, the demand led to the reduction in prices for computing resources, allowing more people to be able to afford buying computing hardware that was prohibitively expensive. Algorithms existed, but the resources were finally able to execute them in a reasonable time and cost.
- Programming languages and packages: Communities such as R and Python developers seized the opportunity, and individuals started releasing packages that exposed their work to a broader community of programmers. In particular, packages that provided machine learning algorithms became immediately popular and inspired other practitioners to release their individual code repositories, making platforms such as R a truly global collaborative effort. Today there are over 10,000 packages in R, up from 2000 in 2010.