This article was originally published on AiThority.
For better or worse, artificial intelligence (AI) is often portrayed as a transcendent solution with the capacity to magically address any business problem. Hyped up AI media coverage has many an executive inadvertently adopting this perspective, seeding the idea that they can reap the benefits of this technology right off the bat —improving their companies’ bottom lines and increasing shareholder value on Day 1 … but when expectation and reality are out of kilter, AI has the potential to generate more problems than solutions.
AI is not new; it has been around for more than 65 years. Since its inception, it has undergone several AI winters, typically characterized by a period of high hopes and overpromises that turn into crushing blows of various degrees when those promises are left underdelivered. Hype cycles are not uncommon with emerging technologies, and those familiar with AI’s history cannot resist pondering if we’re at the foothills of yet another AI winter.
Meanwhile, some executives, especially those who may not have any prior experience with how AI really fits into their business operations, are still trudging through muddy AI waters in the search for quick time to value.
Assessing the AI Landscape
Is another winter coming?
Is AI really the silver bullet that the media is making it to be?
Could it deliver on the promise this time around?
Can it be easily adopted?
Is it still a gamble?
Although these are complex questions with much socioeconomic implication, a frame of reference that gives context to today’s AI promise may shed some light. And while one cannot predict with a great degree of confidence the possibility of another AI winter, some well-established watermarks are worth exploring. So, let us peel away the layers and look at the landscape on a practical level.
Availability of Business Data
One such watermark is the availability of voluminous business data. Imagine the swathes of data that retailers can capture at every step of fulfilling an online order, from browsing and cart-filling behavior to shipping preferences, the resulting warehouse operations and associated delivery sources, routes and timelines. All of this end-to-end business transaction data can be stored inexpensively, indefinitely and with quick accessibility.
Advancement of Machine Learning
Also consider the tremendous advancement of machine learning (ML) algorithms over the past decade and their seamless availability through open-source software channels. Exploratory data analysis, coupled with impressive visualizations, followed by sophisticated deep-learning algorithms combine to help build models that predict outcome at the tips of our fingers; be it product demand or disease diagnosis, these algorithms often exceed human-level performance.
Access to Computational Firepower
Layered on top of this ML is the availability of massive computational firepower through specialized chipsets, such as Google’s TensorFlow Processing Units (TPU) designed to crunch massive model-driven data. These are then stitched together through several Cloud Computing infrastructures (AWS, Azure, GC), where with just a few clicks one can create the necessary computational environments at fractional costs, tackling the most complex data and solution pipelines. Only a few years ago, this would have required large teams and hardware with astronomical costs.
Availability of Consumer-Level AI Solutions
And of course, the availability of disruptive yet transformational AI-based solutions are now available to everyone. Through smartphones and car consoles (Android Auto / Apple CarPlay), one’s traffic-free trips are automatically mapped and constantly adjusting to the unforeseen, while listening to exploratory taste-friendly music; yet another testament that an AI spring, rather than another winter, is more likely.
Assessing Real World AI Compatibility
Certainly, the grass looks green. That said, business leaders should remain wary of hype-driven marketing touting high-powered AI tools. Executives planning to launch AI initiatives as a panacea may find themselves frustrated along their journey, if only because this attitude will often lead to unreasonable expectations, especially if coupled with misconstrued assumptions that an AI investment will follow typical business norms and guaranteed ROI.
Those who are grappling with the practicality and promise of AI may benefit from an “AI reality check” in form of three fundamental questions:
- Is the business problem complex enough that would require the sophistication of AI? (Not all do!)
- Is there enough relevant data to address the problem?
- Is there willingness to experiment?
Only AI-worthy business use cases and healthy doses of data due diligence, coupled with fast-track pilots with adjustment of techniques, technology, expectations and mindset can potentially lead to viable solutions, albeit in a constrained fashion; small AI victories. Building on these victories will undoubtedly lead to bigger successes and long-lasting AI springs. Too big a bite, and there may be undue risk to the business’s bottom line.
Yes, AI itself can be a problem, especially if it is viewed as a solution looking for a business problem, as opposed to the other way around. Even then, it would also be foolish to think that the ever-expanding AI-driven services (such as automated mortgage approvals, employment candidacy validation) will not have impacts on human society, be them ethical, moral, legal and socioeconomic. The full implications of AI-driven complex decision-making are not yet known, and how they may one day impact everyday life should neither be ignored nor underestimated.
AI is a means to an end and not the end itself. It should be viewed as an easily accessible very sharp knife, designed to do precision cutting. If one is not careful and does not handle it properly, they might end up seriously injuring themselves and those around them. But used properly, it’s a powerful tool that can carve out significant value.
