While the authors of the Stanford 100 year study are overall optimistic about the ability of AI to improve society, they caution that special measures may need to be taken to ensure its benefits reach all levels of society equally. They caution that without such efforts, “AI could widen existing inequalities of opportunity if access to [it] is unfairly distributed” [1]. It is certainly possible to envision such a misuse of AI. However, based on the manner in which several other recent technological breakthroughs have proliferated into all levels of society, I believe that this path of stratification is not as significant a concern as the authors make it out to be. 

First, we can look to the recent rise of smartphones as an example of a technology that has provided benefits to all levels of society within a short time of its introduction. Smartphones initially came with a significant price barrier to entry: the average Android phone cost an average of over $400 in 2008. In under 10 years, the mean price of an Android in the US has dropped by over 50% to just $200 [2]. Moreover, in those same 10 years the ever-cheapening smartphone has become nearly ubiquitous, with an 44% [3] of the world population having access to the technology in 2017, compared to just 12% in 2008 [4]. Even within the US, smartphones are increasingly providing opportunities to those with lower opportunities; the percent of low income Americans who use a smartphone as their primary source of Internet access has nearly doubled since 2013 [5]. Certainly, the smartphone started as a tool that served to primarily benefit the upper stratum of society. However, rather than staying concentrated in the hands of the wealthy, the power of this tool quickly diffused throughout society, aided by cheaper and more effective manufacturing techniques. It is conceivable that the rollout of AI may proceed in a similar manner: While the cutting edge of AI may today require massive, proprietary datasets and expensive computers, as the best techniques mature and their behaviors become better studied we will be able to provide them to a larger and larger audience for cheaper and cheaper. 

Even the claim that AI technologies are currently inaccessible to the broader public requires some further examination. Already, a variety of tools exist so that a layperson, or at least anyone with a bit of programming experience, can harness the power of AI. For fans of Python, the free, open source, scikit-learn module [6] provides a suite of easy to use machine-learning techniques for all manner of data analysis. A quick AI-powered Google search for “python facial recognition” yields a modifiable, open source package that can be installed in under an hour and run on a computer as low-power as a Raspberry Pi [7].  Beyond my favorite programming language, excellent resources such as TensorFlow [8] exist with a mission to “accelerate open machine learning research”. Again leveraging the power of Google, one can even find once-price-restrictive texts on the subject of Artificial Intelligence at a fraction of their former prices [9]. 

Overall, while the possibility that AI will lead to the further stratification of society is certainly non-negligible, trends in both other recent technological breakthroughs and the already large availability of open-source AI tools suggest that the technology is already headed in the other, more optimistic direction. 
References:
[1] https://ai100.stanford.edu/sites/default/files/ai100report10032016fnl_singles.pdf, p. 43
[2] https://www.statista.com/statistics/612937/smartphone-average-selling-price-iphone-and-android/ 
[3] https://www.strategyanalytics.com/strategy-analytics/blogs/smart-phones/2016/12/21/44-of-world-population-will-own-smartphones-in-2017#.WbcvkReQxCU 
[4] https://www.statista.com/statistics/218532/global-smartphone-penetration-since-2008/ 
[5] http://www.pewinternet.org/2015/04/01/chapter-one-a-portrait-of-smartphone-ownership/ 
[6] http://scikit-learn.org/stable/
[7] https://github.com/ageitgey/face_recognition
[8] https://www.tensorflow.org/
[9] https://www.google.com/search?q=Artificial+Intelligence%3A+A+Modern+Approach&ie=utf-8&oe=utf-8, check a few links down

