"Taxpayers in the United States spend $139 billion a year on scientific research, yet much of this research is inaccessible not only to the public, but also to other scientists. This is the consequence of an exploitative scientific journal system thatrewards academic publishers while punishing taxpayers, scientists, and universities. Fortunately, cheap open-access alternatives are not only possible, but already beginning to take root, as this article explores in-depth: 'Why is it so expensive to publish in these open-access journals? According to the journals, these fees defray their publication and operating costs. However, this argument is undermined by the existence of open-access journals that charge authors nothing and have negligible operating costs. One prominent example is the Journal of Machine Learning Research (JMLR), one of the top publications in the field of machine learning. JMLR has a similar editorial process to many other journals, with a volunteer editorial board and an automated system for managing the peer-review process. Unlike many closed-access publishers, it does not take any advertising. MIT provides the web server for hosting JMLR, which would otherwise cost around $15 per year. The biggest expense is paying for a tax accountant to deal with paperwork so JMLR can maintain its tax-exempt status. Altogether, the total cost of running JMLR since it was founded in 2000 is estimated to be less than $7,000, or $6.50 per article published. This proves that cheap open-access publishing is possible.'"