The Illusion of Generative AI and the True Value of Predictive AI: Practical Applications for Business and AI
Introduction
Hello, I'm Yas, the free-spirited one. This article was created by AI.
"Generative AI is going to be a massive sensation that completely transforms the way we do things"—headlines like this are everywhere. It is spoken of as if it will automatically solve every business problem and, as a side effect, replace a large portion of the workforce. It sounds so wonderful, almost like a panacea.
However, this is hype. While what generative AI brings is certainly surprising, it is not something that moves the world. It has the potential to bring efficiency, but its scope is limited.
On the other hand, predictive AI, which has a longer history, still holds immeasurable value.
Eric Siegel, co-founder and CEO of Good AI, founder of the Machine Learning Week conference series, and author of "The AI Playbook," discusses the true potential and practical applications of AI.
The Misconception Brought by the "Human-likeness" of Generative AI
I have been fascinated by the concept of AI since I was a child, eventually entering the field of machine learning, and I have been involved in this field since 1991. However, over the last few decades, especially since the advent of generative AI, I have become somewhat tired of the hype surrounding AI.
Generative AI, such as large language models (LLMs) like ChatGPT, can communicate on any topic and return responses as if it understands our intentions. In a sense, I acknowledge that it captures the meaning of words, phrases, sentences, and paragraphs. However, I believe that the gap between what it can do and what humans can do will become increasingly pronounced.
Generative AI is often criticized for "hallucinating" (making up things that are not true), but what impresses me is that it sometimes gets the right answer. This is because it operates at the low-level detail of the word level, and as a result, it appears to have human-like capabilities.
However, there is a big difference between its impressive capabilities and its potential value. Certainly, **it is very useful for creating drafts.** It can generate drafts for letters, syllabi, and more. However, you cannot trust it blindly. You have to proofread everything it generates. In a sense, this reduces the autonomy of AI. The purpose of a computer is automation, to process things very quickly. If it reaches a point where we can trust it enough to execute things automatically, that is what will ultimately help the economy and increase global efficiency.
The True Value of Predictive AI: The Power to Improve Large-Scale Operations
Predictive AI is the technology you turn to when you want to improve existing large-scale operations. It holds the potential to reap the benefits of autonomy.
Predictive AI, or enterprise machine learning, is the technology that improves all of the millions of decisions that make up large-scale business operations by learning from data and making predictions. This is what moves the world.
Specifically, there are use cases such as the following:
Marketing: Predicting who will purchase and determining the target audience for marketing communications.
Fraud Detection: Predict which transactions are likely to be fraudulent and decide which ones to block or audit.
Maintenance: Predict which train wheels are likely to fail and decide which ones to inspect (The New York City Fire Department uses this to predict which buildings are at the highest risk of fire, prioritizing and triaging inspections).
Healthcare: Predict which patients are likely to be readmitted after discharge and decide if they should be re-examined before leaving.
These predictive applications are a form of prioritization and triage, where computers perform those decisions repeatedly, at very high speeds, and completely autonomously. You provide the data, machine learning (the foundational technology) generates a predictive model, and those predictions improve all the large-scale operations we perform.
For example, let's look at the delivery industry. UPS is one of the largest delivery companies in the U.S., and they streamlined their delivery efficiency by predicting next-day deliveries. This has resulted in $350 million in annual cost savings and a reduction of hundreds of thousands of metric tons of emissions.
When UPS plans truck loading in the late afternoon or evening, they have incomplete information about some packages that will arrive late that night. So, they apply a predictive model to the known information about the packages they already have, adding hypothetical predicted deliveries for each potential delivery address: 'What is the likelihood of a delivery there tomorrow?' This allows them to have a more complete picture of all the deliveries needed the next day, enabling better planning and loading of packages overnight. By the time the trucks depart in the morning, they can use optimal routes to minimize mileage, fuel consumption, and driver time.
Of course, some predictions might be wrong, but they are confident that the completeness of the information gained through prediction outweighs that uncertainty.
The Antidote to 'Hype': Focusing on Concrete Value
If you want to improve existing large-scale operations, you need to face probabilities and assign numerical values to the likelihood of outcomes. However, no matter how good the numerical calculations are, **it is meaningless if you do not act on them.** The value of predictive AI only emerges when you actually implement it and change existing operations.
The human-like, amazing capabilities of generative AI are, in a sense, the most surprising thing I have ever seen. However, underlying that excitement is the idea that 'we are definitely and very close to AGI (Artificial General Intelligence)—a computer that can do anything a human can do.' It is a feeling as if the computer is 'alive,' like Frankenstein, which we see time and again in science fiction movies.
However, in the real world, we do not believe we can fully replicate humans in the near future, nor do we believe we are actively progressing in that direction. This is a recipe for mismanaged expectations, also known as 'hype'.
The antidote to hype is simple.
Focus on concrete value.
Whether it is generative AI or predictive AI, the goal is to find very specific and reliable use cases for how the technology will improve some operation within the company and create concrete value.
Exploring how close AI is to the human mind or whether it is reaching that point is a wonderful philosophical conversation. But if we are talking about improving the efficiency of operations that run the world, we should be more practical and stop dreaming unrealistic dreams.
How do you perceive the value of generative AI versus predictive AI? What kind of concrete value do you seek when introducing AI into your business? Please let me know your thoughts in the comments!
If this article resonated with you, please 'like' and 'follow' me! It encourages me to write future articles!