Artificial Intelligence and Competitive Advantage (An Introduction)

How does a company stay ahead with Artificial Intelligence?

With the Artificial Intelligence industry arguably just emerging in the last decade or so, one can’t help but wonder how the true value of AI can be conceptualized. In other words, can we easily state that firms that work with Artificial Intelligences, to some extent, have some sort of competitive advantage over those that don’t? First of all, plenty of speculation as well as serious research has been done on this subject. Different writers and researchers have taken different stances on this with some sources stating that AI can help with overall marketing, including social media marketing, and others stating that AI even helps with sales as well as overall business intelligence. There does seem to be a consensus, however, among various sources, that the biggest advantage related to using AI is what it can do with data.

Whether this means data governance and security or just “crunching the numbers,” AI has proven time and time again that it can work with data in quicker and more efficient ways than humans can alone. If you don’t believe me, just look at the importance of Big Data, simply via the sea of consultancy reports that are still coming out on the subject. If you’re not already clear on what “Big Data” is or how we can best understand it, think about this. Companies often use extremely large sets of numbers and facts about customers to learn more about them and drive future product offerings and therefore, sales. The data sets that are used in this specific way can be thought of as “Big Data.”

Big Data can be analyzed by humans alone. That’s why we have certain programming languages and that’s one of the Python language’s biggest uses. There are also other languages like R, which some seem to believe were created almost specifically for data analysis.

In theory, yes, AI can do everything related to this, faster than humans can.

It all begins with human limitations. Humans have to take breaks, including for sleeping. Artificial Intelligences can work around the clock and can, in theory, process more data or simply do more calculations than the human brain could ever dream of doing.

Limitations of AI

Even so, critics like McKinsey and the South China Morning Post have pointed out that AI systems are still only able to do as much as they are trained to do. In other words, take a look back at our Dbrain piece. As of now, any sort of data that an AI system works with or will work with, must be labelled before it is used in practice. That means that if it needs to consistently identify certain animals in certain pictures, then the math that leads to the correct conclusions for this needs to be input into its system by human programmers, essentially from the inception of the AI system. This may not seem so bad, but when this point is added together with the point that the larger the training set, the more accurate the AI system consistently is, then everything can seem a bit more troubling.

Here, by training set, we mean that data that has to be input from the very beginning of an AI system’s “life,” and so, as critics like McKinsey have noted, this raises problems related to determining what the ideal amount of data is.

Furthermore, what is the ideal level of accuracy? It varies from one AI system to another but in general, it’s best to think of this answer as being: the higher, the better. Even now, AI researchers and scientists are working towards systems that will consistently work in as flawless a fashion as possible.

Where does this leave us?

All in all, this leaves our principal question unanswered. How can AI bring competitive advantage to a business beyond streamlining processes, processing more data and possibly finding new insights through this data processing ability? AI Business, which is a news portal that claims to be the top spot for all AI news, appears to state that we can increase the possibility of an AI-related competitive advantage via using our data in two specific ways.

These are: Data Governance and Data Cataloging or Labeling.

The argument here is that data governance or the practice of developing and using a strategy to determine how a firm will determine what data is ideal and what data is not, will increase trust in data and therefore in new solutions related to dealing with it. The author of this particular piece then suggests that Artificial Intelligence systems are the principal solution that can be considered in this fashion in that they also often predict future problems related to the data that they work with, well in advance. In terms of Data Cataloging, we mean identifying what the ideal values are for a system, through training sets as in our Dbrain piece. In other words, through teaching the system, we learn what it really “needs” to be finding and how it needs to be finding it.

When these two practices are implemented well, it is suggested that trust is increased to the point that such a business feels secure in examining the viability of new solutions, like allowing an AI system to control its data.

Furthermore…Machine Learning

Once this is done, it’s also implied that Machine Learning is the actual catalyst for creating a competitive advantage with AI. In stating this, it’s also claimed that since Machine Learning involves computational methods and algorithms, which are the factors that run an AI “brain,” then without the existence of Machine Learning, no AI systems would exist. In giving an AI and therefore, its machine learning algorithms, full access or even close to full access of a firm’s data, concerns arise related to security.

No system is perfect.

It is entirely possible that an AI could slip up one day and accidentally release sensitive data to some sort of public source. This is yet another reason why data governance before even considering working with AI is essential. A firm needs to have clear rules as well as developed scenario analyses so that it is perfect sure how it will act in any situation that’s anticipated to be a high-risk. Going back to Machine Learning as the catalyst of competitive advantage, it could be argued that its true utility is in its data crunching ability. Some sources report that efficient AIs can run through billions of records, find common themes and make conclusions based on these themes. One specific example that is mentioned in the AI Business Blog is the capability of such a system to analyze large numbers of medical records, no matter the volume, and conclude with possible diagnoses related to the medical histories of individuals.

We all know that paperwork of any kind as well as document analysis of any kind is time consuming and takes us away from higher-level, cognitive work. With that in mind, this seems truly exciting that AI can take the pressure off of individuals like doctors, in this way. Despite this, we should still remember that AI systems are not without their limitations. They need to be trained and that takes human hours. Furthermore, they need to be explicitly accurate, especially when the success of a product hinges on their results or even more importantly, people’s health hinges on their results.

With these yet to be solved weaknesses in mind, it will be interesting to see how the Artificial Intelligence industry decides to move forward towards a yet uncertain future.