Written By: CENSA Editorial Board
For many years the greatest artificial intelligence (AI) minds toiled without fanfare in the academic world while commercial breakthroughs and even more basic applications of AI research remained unrealized. Yet with increased investment from the likes of Google, Facebook and Amazon, with the availability of open source software frameworks such as TensorFlow, and with an upsurge in the number of niche, AI-focused startups, an explosion of AI-related products and services has recently hit the marketplace. For AI proponents, it is truly an exciting time.
AI has the potential to completely transform the way work is performed in the world today, especially in highly structured industries. Many startups understand this and have already begun to demonstrate potential capabilities in so-called “grey markets,” or sectors of the business world that are less regulated and ripe for low-profile experimentation. Designed and executed discreetely in order to avoid an increase in regulatory oversight and action, these efforts offer practical innovators the opportunity to promote and expand the art of the possible in AI. Yet grey markets might prove to be even more important for policy makers to follow and understand because as these products gain more traction and usage they are likely to shape, define, and foreshadow the limits of future policy options.
The development and rise of artificial neural networks (ANNs) is an example of an AI-innovation that is widely used to enable machine learning. Based on what we understand to be the workings of the central nervous systems of animals (with respect to the brain), and coupled with advanced mathematical models, ANNs have been employed by Google and Facebook to sort through and analyze large data sets, a process through which multiple tasks can be executed. In traditional programming, a software engineer will write code containing specific instructions and parameters for a computer to follow. Yet with AI-induced machine-learning the underlying code remains constant and unchanged, even as the computer continues to adapt and find new ways to accomplish specific end states (i.e., as it “learns”). In these circumstances a software engineer has only a broad sense of what might occur and has only a general idea about what might cause a computer to behave in a certain way. This so-called “literacy gap” among computer designers and operators is important to remember, even as the number of fully independent machine-learning systems increases, because it suggests that the workforce must learn to adopt a more flexible mindset (and perhaps new skillsets), and accept a new level of uncertainty with respect to machines.
The great potential of AI products actually lies in the ability of the designer and user to anticipate different aspects of their potential, and to manage, leverage, and utilize them to make future tasks even more productive and efficient. Many of these systems, though flawed, possess not only the ability to grow more sophisticated but the ability and potential to produce unexpected and unfortunate outcomes. The Microsoft Tay incident of March 2016 is a case in point, when the chatbot named Tay posted several inflammatory, offensive, and unfortunate statements online. In accordance with its underlying instructions to adapt to the communications’ environment, Tay simply mimicked and countered and then intensified (and inflated!) the inflammatory slang it encountered. The incident is an excellent reminder that even seemingly harmless AI-based creations can produce unintended side effects and negative results.
Trends associated with the sudden surge of commercially available AI products and services will continue to develop as the technology advances further and industry attempts to scale and grow various products and services. For policy makers – especially those with an interest in (and a keen eye set on) national security matters – the evolution of grey markets, literacy gaps, and unintended effects will prove to be key. Understanding their development and how industry reacts and responds, must inform and fundamentally influence government action.
 Jack C. (2016, May 18). Google Shows Off to Keep Up with Facebook, Amazon, Apple [Online News]. Retrieved from http://www.bloomberg.com/news/articles/2016-05-18/google-shows-off-ai-to-keep-up-with-facebook-amazon-apple
 Lisa C. (2016, April 12). 15 Game-Changing Artificial Intelligence Startups [Online News]. Retrieved from http://www.inc.com/lisa-calhoun/see-13-of-the-artificial-intelligence-companies-checking-you-out-today.html
 Shivon Z. (2015, December 10). The Current State of Machine Intelligence 2.0 [Online Article]. Retrieved from https://www.oreilly.com/ideas/the-current-state-of-machine-intelligence-2-0
 Cade M. (2015, May 6). Building AI is Hard – So Facebook is Building AI That Builds AI [Online News]. Retrieved from http://www.wired.com/2016/05/facebook-trying-create-ai-can-create-ai/
 Jason T. (2016, May 17). Soon We Won’t Program Computers. We’ll Train Them Like Dogs [Online News]. Retrieved from http://www.wired.com/2016/05/the-end-of-code/
 Cade M. (2016, June 21). Forget Doomsday AI – Google is Worried About Housekeeping Bots Gone Bad [Online News]. Retrieved from http://www.wired.com/2016/06/forget-doomsday-ai-google-worried-housekeeping-bots-gone-bad/