August 20, 2020. Steve Shillingford

Moving Beyond the Hype: Implementing AI to Solve Real-World Business Problems.

Many of us are all too familiar with the infamous “Gartner Hype Cycle.” The model is a branded version of other models to represent the maturity, adoption, and social application of specific technologies. The hype cycle claims to provide a graphical and conceptual presentation of the maturity of emerging technologies through five phases. Each hype cycle drills down into the five key phases of a technology’s life cycle.

No.
Phase
Description
1
Technology Trigger
A potential technology breakthrough kicks things off. Early proof-of-concept stories and media interest trigger significant publicity. Often no usable products exist and commercial viability is unproven.
2
Peak of Inflated Expectations
Early publicity produces a number of success stories—often accompanied by scores of failures. Some companies take action; most don’t.
3
Trough of Disillusionment
Interest wanes as experiments and implementations fail to deliver. Producers of the technology shake out or fail. Investment continues only if the surviving providers improve their products to the satisfaction of early adopters.
4
Slope of Enlightenment
More instances of how the technology can benefit the enterprise start to crystallize and become more widely understood. Second- and third-generation products appear from technology providers. More enterprises fund pilots; conservative companies remain cautious.
5
Plateau of Productivity
Mainstream adoption starts to take off. Criteria for assessing provider viability are more clearly defined. The technology’s broad market applicability and relevance are clearly paying off. If the technology has more than a niche market then it will continue to grow.[1]

 

[1]https://en.wikipedia.org/wiki/Hype_cycle#cite_note-1     https://en.wikipedia.org/wiki/Hype_cycle

Like gravity, the hype is everywhere:

  • At the start of 20th century, there were over 100 car companies in the United States.
  • Now there are 2. (Chrysler is owned by an Italian company)
  • By the mid-1950s, there were 10,000 radio manufacturers and 500 television set makers. Now there are less than 10 each in the world.
  • More recently, there were over 500 companies known for making personal computers; today there are five.
  • In the 1980s, there were over 30 video game consoles. Now, there are 2. If you count the Switch, 3 (but I’d say 2 and half)
  • We’ve seen over 100 different mobile operating systems. Can you name one that isn’t IOS or Android?

Finally, remember when we needed to know our credit card number to order online? Now, it’s “1-click” to shopping Nirvana, courtesy of Jeff Bezos.

I am a big believer in the Hype Cycle as a framework for understanding patterns in technology. We can see these cycles everywhere and while the chief critics complain about the lack of “scientific” evidence for its existence, there’s no denying its veracity.

Rapid, almost Cambrian-like explosion in innovation, followed by failed promises, disappointment, and cynicism. Then, consolidation and eventual broad-based adoption driven by the remaining winners.

With this in mind, we are predicting the next wave of consolidation to occur in the overheated market of Data Science.

Irresistible Force Paradox: Theoretical meets Tactical

For clarity, we’ll stipulate the differences between “Artificial Intelligence” and “Data Science” here.

Data Science is the analysis and study of data. Many believe a Data Scientist is responsible for making decisions that benefit companies. Moreover, the role of data scientist varies with the industry. In the everyday roles and responsibilities of a data scientist, the main requirement is to preprocess data, that is, performing data cleaning and transformation. He then analyzes the patterns in the data and uses visualization techniques to draw graphs that underline the analytical procedures. Then he develops prediction models that find the likelihood of the occurrence of future events.

Artificial Intelligence (AI) is a tool or a procedure. This procedure sits at the top of the other methodologies, used for analyzing the data. We should think of AI as a tool leveraged by data scientists for creating better products and imparting them with autonomy in hopes of automating, emulating, or improving human decision-making.

While distinct, it is important to remember that data science is a practice and AI is one of the many tools deployed by said data scientist.

We will use these terms interchangeably, but what is important for the reader is to remember that the evolution of decision making, informed by the recent organizational promotion of data scientists to high priests of decision-making, has brought “AI-powered” solutions to the forefront of every C-level executive’s desk.

In the last 5 years, we’ve seen an explosion of innovation in the space. New ways of analyzing old data have certainly brought insights into how models can offer value, but many operators will privately concede that their investments in alphabet soups like AutoML, AIOps, RPA, and NLP have not paid off. A recent article in MIT Technology Review summarizes this nicely:

To quote a classic paper titled “Machine Learning that Matters” (pdf), by NASA computer scientist Kiri Wagstaff:

Much of current machine learning research has lost its connection to problems of import to the larger world of science and society.

https://www.technologyreview.com/2020/08/18/1007196/ai-research-machine-learning-applications-problems-opinion/

This is not the fault of Data Scientists, by the way. In fact, most job titles with the word “scientists” in the name are generally immune to the pressures of profitability, risk, or operational improvement. Rather, the title implies doing science: Forming a hypothesis, conducting experiments to prove or disprove, rinse and repeat.

In 2020, even the most profitable businesses can no longer afford science projects. Massive, tectonic, and global shifts in work forces, consumer demand, and public health concerns have forced companies to not only cut, but to also accelerate the adoption of any and all technologies that reduce costs, mitigate risks, and improve customer satisfaction. The new normal of remote work and decentralization means looking everywhere for solutions that promote real, tangible, business returns. In a world where data science was viewed as a theoretical exercise, decoupled from outcomes, we are seeing line of business executives exert pressure on their data scientists with a sentiment that could be described as “what have you done for me lately?”

So, in a market where the hype has peaked, the choices for various tools is overwhelming, and the immediate needs of business-level return on investment have never been greater, what to do?

The answer? Like Captain Ramius from the The Hunt for Red October, you steer directly into the torpedo!

The market for AI is not immune to hype. When I was growing up in the 80s, we were told flying cars were right around the corner. In the 2000s, we were told cars would drive themselves. Today, I settle for an effective cruise control…which is fine, unless, of course you’re one of those that trusts a Tesla not to speed directly into a bridge abutment. Tesla likes to pretend their cruise control is actually “semi-autonomous” car driving; that this is premier AI in action. If you’re the guy who burned up as a result of a bad decision, you might think otherwise. But that’s not the mark of a peak. It’s when my favorite golf club manufacturer leads their Sunday commercials with “AI design” drivers. I would pay extra for AI that fixed my snap-hook, but let’s face it, that doesn’t exist. Much less dangerous, but over-hyped just the same.

We’ve just spent the last 5 years witnessing the Hype Cycle again in the AI space.

https://www.gartner.com/smarterwithgartner/top-trends-on-the-gartner-hype-cycle-for-artificial-intelligence-2019/

According to insights from Gartner, there are many emerging technologies and trends in the AI space, but few offer value or complete understanding to drive business results.

There are a number factors driving this market’s hype:

  • Advances in AI research, especially open-source innovation, has increased in both volume and velocity in recent years.
  • The AI product market has reached peak hype driven by a failure to deliver on business requirements while driving increasingly specialized niche solutions.
  • The AI marketplace for data scientists, and the forecasted “lack of talent” has not come to pass.
  • Current macro-economic conditions and globally decentralization efforts are driving increasing demands for “AI-powered automation.”
  • History predicts a consolidation around platforms that support open innovation while solving the friction and role integration challenges between those that manage the models and those that manage the business.

Bottomline: Data science, like math, is hard. Numbers are discrete but reality is messy. While the beauty of the models and all the innovation are exciting, there is still a long way to go between the inherent beauty of a model and its usefulness in any setting, let alone business. In the words of Ramaz Naam:

In the end, I expect we’ll have AI that is better than we are at nearly every narrow task but which are still our tools, not our masters.

And therein lies the problem: We outsourced the power of AI to scientists. Smart in their own right, connecting the dots between the science of data and the context of decision-making is hard for anyone. Even a scientist. As the masters of AI, it’s up to us to operationalize all the goodness that data science offers and today’s niche tools, technical platforms, and host of erector-set like pieces are geared toward helping scientists do more science. Hence, there’s a reason why we have data scientists and data managers in separate roles in the enterprise. And that separation is part and parcel to the misfit between data science investments and data science returns.

It’s time to fix this.