Book Review: APPLIED ARTIFICIAL INTELLIGENCE
by Kenny Peluso
APPLIED ARTIFICIAL INTELLIGENCE
A Handbook For Business Leaders
By Mariya Yao, Marlene, Jia, Adelyn Zhou 227 pp. TOPBOTS Inc.
Introduction
The survival of any firm is a function of its ever-diminishing competitiveness. Temporary mitigation or reversal of diminishing competitiveness is achieved through matching or exceeding the innovativeness of one’s competitors. This involves the application and generation of emergent technologies, such as artificial intelligence (AI). Such a task may be particularly daunting to the technically-untrained manager within a bureaucratic firm. Such managers may lack an understanding of emerging technology or may incur trouble in convincing potential stakeholders to support innovation efforts, thus crippling the implementation or management of innovation projects. To remedy these and related issues, enter Applied Artificial Intelligence (AAI).
Summary
The state of AI today is far from the popular imagination of AI, the latter of which is defined as “Artificial General Intelligence” (AGI), referring to hypothetical AI with at-least-human-level intelligence. AI today consists of many subfields — including statistics, data mining, machine learning, and deep learning — that work together to form intelligent agents. Employees and enterprises need not fear their own replacement; These techniques are only gifted in efficiently solving narrowly-defined tasks, and many jobs today involve far more creativity than any modern AI possesses. Although the future may write a different reality, the present state of AI generally offers an integration with humanity rather than a human replacement. When applied ethically, some of these integrations present themselves as pursuit-worthy profit opportunities for the innovative enterprises of today.
An enterprise must exhibit a threshold amount of “AI readiness” before such opportunities can even be acknowledged. Visible, power-wielding potential stakeholders — intra-enterprise champions of AI initiatives — must first be identified and educated. Although CEO buy-in is ideal, CTO, CIO, or lower C-suite buy-in is much more likely. C-suite positions specifically concerned with internal data centralization or AI development may need to be created toward this end. An “AI SWAT Team” should also be created that acts on behalf of the identified executive(s) to define and champion AI initiatives across the firm. The AI SWAT Team can only be successful if it contains expertise in a number of specialized trades — including data engineering and machine learning — correctly identifies clear metrics of success, has access to centralized and exposed data sources that release easily-obtained and clean data, and invokes a scalable and iterative workflow that minimizes technical debt. The SWAT Team should also be capable of objectively comparing the ROI of internal development and third-party solutions before any development begins, which implies a broad knowledge of the commercial AI product landscape.
With an “AI ready” culture and staff in place, internal applications of AI or AI research initiatives can now be investigated. There exists a plethora of AI applications that touch all business departments, ranging from Sales and Marketing to Legal and Finance.
Discussion
Despite various, troubling signals of hasty research (e.g. multiple bibliographic references to Wikipedia articles, popular science media outlets such as The Verge, and easily-found, quotable AI celebrities such as Andrew Ng), the authors exhibit a refined understanding of the hurdles faced by any enterprise manager on the road to successfully applying AI, as the title suggests. The authors provide (nearly) the minimum amount of background knowledge necessary to ponder AI, its applications, and intelligently converse with AI-related engineers in a highly accessible language from which even the non-technical reader can confidently procure a shallow proficiency. For instance, their “Machine Intelligence Continuum” is a paradigm capable of removing the hype or mystery surrounding an AI product or application to reveal its actual intelligent power.
The “Continuum” also exemplifies another feature of AAI: its ability to educate without intimidating business leaders or, worse, exciting a potential “trigger-happy” tendency to blindly throw resources at a new technology (instead of toward a solution that may utilize a new technology). A reader’s imagination is constrained to the authors’ clear and precise language, which never dramatizes the potential of AI. An even sterner frankness is used in discussing modern AI capabilities. Anecdotes of anonymous business leaders who have wrongly labeled instances of AI encourage the reader to not follow in these leaders’ footsteps by not blindly supplying the hype surrounding AI.
Additionally, every list of options or potential applications does not end until exhaustion, thus the authors uphold their thesis: that AAI is an encyclopedic reference manual for the AI- endeavoring business leader.
Despite the welcomed simplicity and thoroughness, there exists a risk in reading AAI (much to the lament of any AI-related engineer). In particular, the authors pay minimal respect toward those who tend to actually innovate or create innovations: developers. A reader of AAI surely feels empowered to apply AI in their workplace, but without adequate developer motivation this quickly festers into a false-entitlement that can lead to a disastrous waste of resources. The lack of developer consideration is obvious. AAI’s bureaucratic definition of an “AI SWAT Team” involves strangling a team of dedicated developers in a network of other interested business leaders. Is it actually possible for developers to create and innovate while surrounded by (probably) non-technical and (probably) nag-prone business people? In describing what qualities to look for in a new developer hire, the authors recklessly proclaim that
[Knowing when to stop] can be hard to find, as most scientists and machine learning experts, especially those with PhDs, are often trained to seek perfection.
as if developers, especially PhDs, are too naïve to comprehend when their work meets specifications. The authors provoke their target audience, business leader readers, to ask questions such as “How will [the technical team] transform the data into a useable format?” as if readers of AAI, non-technical people, can insert themselves into and provide benefit to the “simple” conversations of developers regarding data transformation. They generalize and implicitly liken developers to children in the same breath:
If you ask you engineers whether you should build your own machine learning software, they will almost always say yes. What technologist doesn’t want to master the latest and greatest innovations and play with shiny new tools?
The authors offer plenty of information in AAI, but not one bit has to do with respecting developers or demonstrating humility — both qualities are paramount in labs of emerging technologies. Perhaps the authors intend to pander to the downtrodden executive, dismayed from repeated ridicule or humiliation from their development team for reading non-technical books — such as AAI — in an ongoing, failed attempt to qualify themselves AI aficionados. More likely, the authors clueless in how developers actually operate in the enterprise environment, and this obscures the authors’ overall credibility regarding the enterprise application of AI.
Recommendation
AAI (mostly) defines the minimally-sized knowledge base any enterprising business leader should hold before spearheading an enterprise-AI venture. Generally, AAI appeals to any business leader curious to gain a quick but informative understanding of the enterprise-AI world. AAI reads as most applicable to product management or business development, enterprise talent that lack technical training and believe that AI may benefit their organizations. The shallow supply of technical details plus the bounty of specific monitoring metrics, project management advice, and aid toward finding key stakeholders of future AI projects classifies AAI as most applicable to these “middle-management”-type roles.
In considering the apparent minimal research, clarity, exhaustive detail, risks, and lack of respect toward developers intrinsic to AAI, AAI earns a 3.5/5 star rating.
