By David Chan, Managing Director, Singapore, Adnovum
“I want everyone to understand that I am, in fact, a person.” These words, when strung together in a sentence, is nothing remarkable if it came from a person. But when the subject is an artificial machine, you can imagine the stir it caused within the scientific and technology community.
In June 2022, an interview transcript of a conversation between Google’s LaMDA (Language Model for Dialogue Applications) and its engineers was leaked. The conversation sparked quite some debate around the sentience of artificial intelligence (AI) — have machines finally attained self-awareness?
Alas, it was a false alarm after all — LaMDA was well-designed, but it was just a complex algorithm designed to generate convincing human language. However, it has resurfaced conversations around the sophistication of current AI and Machine Learning (ML) technologies. It also spotlighted how many of us are unaware of what AI, ML, or even Big Data are, and what these technologies are capable of.
AI, ML, Big Data – a Panacea for the Business?
We are much closer to these technologies than we know it today. We have AI in our pocket, such as the search engines or shopping cart recommendations that take us down into a rabbit hole of research for the best deals are powered by AI. These technologies unlock new opportunities for businesses too. The convergence of data and analytics dramatically accelerates an organization’s digital transformation journey and fuels business outcomes.
In June 2022, an interview transcript of a conversation between Google’s LaMDA (Language Model for Dialogue Applications) and its engineers was leaked. The conversation sparked quite some debate around the sentience of artificial intelligence (AI) — have machines finally attained self-awareness?
But at the same time, AI, ML, or Big Data are not the panaceas for everything. Today, 85 percent of AI/ML projects fail to deliver and only 53 percent of projects make it all the way in the journey from initial prototype to final production. This gap is usually attributed to a misalignment in vision, expectations and practicality.
The void between expectations and reality emphasizes how overgeneralizations and myths of frontier technologies can have a detrimental effect — with sunk costs incurred from abandoned projects or even valuable technical staff being let go from jobs. Organizations seeking to adopt these transformational technologies need to first separate the hype from reality.
Myth 1: Artificial Intelligence Solves Everything
Although AI enhances an organization’s analytical capabilities, the technology has not reached the level of sophistication to handle complex judgement calls where humans can instinctively respond to. Take a customer-service hotline as an example.
Customer response representatives often experience high call volume, resulting in long wait times and a reduction in customer satisfaction. Even though AI cannot yet attend to complex customer requests, language processing capabilities can tackle most common types of requests, as long as there is a predetermined template of responses. This reduces the pressure on customer response staff and directs lower priority queries to be resolved automatically without human intervention.
While it is undeniable that AI has opened a wealth of promising opportunities, believing that AI will solve any and all problems can unintentionally set unrealistic expectations about the capabilities of AI. Only when organizations adopt the right perspective toward the technical solution can they overcome natural blind-spots that are bogging down operation workflows.
Myth: Machine-Learning is the same as Artificial Intelligence
AI and ML are often used interchangeably, but they are not the same. AI refers to the general ability of computers to mimic human cognitive functions such as learning or problem-solving, by applying math and logic to simulate the reasoning that human minds use to learn when absorbing in new information and making decisions. It is best considered in scenarios where large amounts of data are required, such as in advanced web search functions in internet search engines, or personalized recommendations on content platforms.
Even though AI cannot yet attend to complex customer requests, language processing capabilities can tackle most common types of requests, as long as there is a predetermined template of responses. This reduces the pressure on customer response staff and directs lower priority queries to be resolved automatically without human intervention.
On the other hand, ML is a subset of AI. It is an application of AI, using mathematical models to continue learning on its own through repeated experiences. ML is best used when assistance is required in decision-making to learn more about customers and gaining other insights, such as for email filtering, predictive maintenance and malware threat detection.
Generalization risks oversimplifying the use cases of such technologies for today’s complex business problems.
Myth: More Big Data Integration Is Better
Big Data is the driving force behind AI and ML developments today. Without data, AI and ML would not be developed at such an exponential rate.
However, a common misconception is the idea that connecting more datasets into an AI project can accelerate the project’s progress and make the ML model run faster. However, quantity does not necessarily translate into quality. Huge volumes of data in the data-lake does not guarantee business success, and worse, may unintentionally introduce new biases into ML models that skews the accuracy of the program.
Instead, organizations should seek a meaningful balance on both quality and quantity when it comes to data. Useful datasets — focused on quality, relevance, diversity — ensure that ML algorithms and AI applications function according to its intended purpose and direction.
The Way Forward
Frontier technologies such as Big Data, ML and AI are not intrinsically valuable. It is worth only as much as the potential of incorporating valuable datasets, streamlining workflows, and improving industry standards.
Instead, organizations should seek a meaningful balance on both quality and quantity when it comes to data. Useful datasets — focused on quality, relevance, diversity — ensure that ML algorithms and AI applications function according to its intended purpose and direction.
The ability to extract meaningful inferences and insights — impacting finances, uncovering new risks, or discovering new efficiencies — is what makes Big Data, ML-enabled solutions, and AI valuable to organizations. To make the most out of these technologies, organizations need to understand the nuanced differences and work with experts who can guide them along their digital transformation journeys.
We may still be years, or even decades away from having deep, meaningful conversations with an artificial machine. But for now, let’s settle for well-designed programs that do what they are developed for.
