GenAI Gets the Spotlight, But Predictive AI Does the Heavy Lifting

"One day the AIs are going to look back on us the same way we look at fossil skeletons on the plains of Africa: an upright ape, living in dust with crude language and tools, all set for extinction." - Nathan, Ex Machina (2014)

Putting aside the AI-generated transition videos and the deepfake clips of renowned personalities giving outlandish advice that pop up on my LinkedIn feed far more often than I would like (perhaps it's time to curate who I follow), Generative AI, or GenAI, is truly amazing; it can create art, write stories, and even mimic human conversations in the voice of Scarlett Johansson ! But is it really the universal AI solution everyone is talking about?

The conversation around AI started picking up after ChatGPT's launch. In fact, that's when AI began to surpass Fortnite in Google searches (for the uninitiated, Fortnite is an online video game). Since then, there has been no Turing back (pun intended). Specifically, GenAI, which is what most people think of when they hear the term 'AI', has garnered significant attention. However, the excitement surrounding it has led many people to view GenAI as a one-size-fits-all solution to business challenges. But in reality, we shouldn't view GenAI as the ultimate answer everyone expects it to be. Instead, Predictive AI provides a more practical value proposition, revolutionizing how companies make decisions and enhance efficiency.

But before diving into the deep end, let's take a step back and briefly understand the types of AI out there.

Different Types of AI

As per IBM's definitions , there are three types of AI based on capabilities:

1. Narrow AI is AI designed to complete very specific actions, often faster and better than a human. It's unable to learn independently. Narrow AI can further be split into two categories, each with distinct functionalities and limitations:

a. Reactive Machine AI is designed to perform a particular task, but doesn't store any memory or learn from past experiences. Examples of Reactive Machine AI are IBM Deep Blue, a chess-playing AI that beat chess Grandmaster and then world champion Gary Kasparov in 1997, and the Netflix Recommendation Engine.

b. Limited Memory AI can use historical and present data to decide on a course of action most likely to help achieve a desired outcome. Unlike Reactive Machine AI, it can learn and improve over time. GenAI is a type of Limited-Memory AI, as are virtual assistants like Siri and Alexa, or self-driving autonomous vehicles. GenAI can further be segregated into subsets, with Large Language Models (LLMs) being one type of GenAI.

2. General AI (AGI), NOT to be confused with Generative AI (GenAI), is just a theoretical concept today. It is an AI that should be able to learn, think, and perform tasks at levels similar to humans. This type of AI could sense human emotions, understand social cues, and engage accordingly with humans - almost sentient but not quite. Estimates vary vastly on when AGI will be ready. Personally, I resonate with what Andrew Ng , one of the most capable AI computer scientists in the world, says: "The standard definition of AGI is AI that could do any intellectual tasks that you can. So when we have AGI, AI should be able to learn to drive a car, or learn to fly a plane, or learn to write a PhD thesis in university. But that definition of AGI, I think we're many decades away, maybe even longer.". So thankfully, there won't be any 'Ex Machina'-esque or 'Roko's basilisk '-type situation occurring just yet. I think there needs to be further advancements in quantum computing and neurotechnology both before AGI comes to fruition, but that's a discussion for another time.

3. Super AI (ASI), which is AI that would evolve beyond the point of understanding human sentiments and experiences. It would have needs and desires of its own. If ever realized, ASI would think, reason, learn, make judgments, and possess cognitive abilities that surpass those of the most intelligent humans. If anyone's curious how AI Singularity would start, it would be with the development of a self-aware ASI.

Returning to our discussion, Predictive AI would historically be classified under Limited Memory AI, which if you remember, can recall past events and outcomes. But its classification is changing as we speak. With advancements in AI in general, Predictive AI is transitioning towards becoming a mix of Limited Memory AI and Reactive Machine AI. But what does that mean? This hybrid approach enables Predictive AI to balance precision, agility in responsiveness, and experience, thus improving its real-time decision-making capabilities - more on this in the next section.

GenAI vs. Predictive AI, Specifically from a Marketer's Perspective

According to Bernard Marr , the author of "Generative AI in Practice ", the key difference between these two types of AI lies in their respective business applications. GenAI is typically used for creativity-driven and research-driven verticles. In creative industries, GenAI has been instrumental in producing original music, designing artwork, and developing scripts. Furthermore, it's even being used for legal research, where it assists in summarizing documents or generating insights. And let's not forget about the inroads made into software development by automating code generation. Here are some additional use-cases actualized by companies using Google's Gemini, along with customizable AI agents available on Google Cloud.

Now putting my marketer shoes on, I can say that GenAI is incredible at creating personalized advertising content. Meta is already taking steps in this direction by launching their proprietary GenAI ad creative tool for video expansion, image animation, and text generation a couple of weeks ago. Additionally, as demonstrated by Google's experiments with GenAI , these tools are helping marketers explore creative possibilities and develop new, engaging content more effectively.

GenAI can also significantly enhance customer experience (as shown in one of Gartner's recent studies ) by enabling tailored recommendations, chatbots for instant support, and customized & interactive content that engages users more effectively. This reflects GenAI's significant potential to drive customer loyalty. By 2025, it is projected that 73% of businesses will utilize AI to manage customer experiences.

However, GenAI's outputs are not always predictable or reliable for operational tasks, as they often lack accuracy and consistency. While OpenAI's GPT-4 did pass the Bar , it still failed to help me prepare for my GMAT (refer to the screenshot below, where ChatGPT answered a Permutation & Combination question incorrectly - one of the many similar challenges I have faced with ChatGPT). It also fails to follow basic chess rules (in a match against a YouTuber) , something we already know a Reactive Machine AI has been able to do quite easily for decades!

Predictive AI, on the other hand, excels at optimizing processes, improving decision-making, and reducing uncertainties in business operations by using statistical modeling & machine learning to conduct data analysis, forecast trends, and anticipate customer behavior. Moreover, Predictive AI is already transitioning to react in real-time by providing valuable insights by both analyzing live data and drawing upon past experiences.

As discussed in Meta's Business Innovation & Technology podcast , and in Deloitte's report on AI's impact on marketing , such advancements in Predictive AI are revolutionizing marketing analytics, enabling marketers to make informed decisions and create highly targeted campaigns that maximize ROI in real-time. If you are not in the marketing, media and advertising industry, let me tell you, THIS IS BIG! Because previously, marketers had to wait weeks before any predictive model could adjust its output in response to changes in any particular data set.

But the emphasis on Predictive AI isn't just about optimizing marketing budget allocations; it's also about driving customer engagement in a targeted, data-driven manner. It also empowers campaigns with greater precision, allowing advertisers to prioritize budgets towards audience segments that are most likely to engage or convert. A typical example would be look-a-like modeling (expanding your reach to a broader audience that shares similar traits with your target demographic), which has been around for ages - on a separate note, this is now being repurposed as Advantage+ Audiences by Meta and Optimized Targeting by Google ('same same but different'-vibes).

Coming back to the GenAI vs Predictive AI discussion, understanding the distinctions between the two is crucial for businesses looking to leverage AI effectively in their operations, especially in the short-term. Both types of AI have immense value, but Predictive AI tends to deliver more reliable and actionable insights, making it more suitable for organizations focused on improving operational efficiencies and making strategic, data-driven decisions.

Before jumping to conclusions though, let me step outside my comfort zone and talk about Predictive AI beyond marketing.

Predictive AI: The Unsung Hero of Business Efficiency

While GenAI dazzles, Predictive AI delivers. Predictive models can enhance large-scale operations by making data-driven decisions that significantly reduce costs and boost performance.

Take UPS, one of the top logistics companies, as an example. By using predictive models, UPS plans its deliveries before all packages are even in its system. This future-facing approach, combined with prescriptive models for optimal driving directions, saves them over $350 million annually while reducing emissions by hundreds of thousands of metric tons. UPS achieves this level of efficiency by augmenting known data of packages already in its possession with predictions of potential deliveries that are yet to arrive. By assessing the likelihood of packages coming in later at night, they get a better picture of the next day's workload. This allows UPS to optimally load trucks and plan delivery routes that minimize driving distances, fuel usage, and driver hours. Furthermore, with an astute analysis of various factors like traffic patterns and weather conditions, UPS also proactively reroutes parcels to avert delays, thereby upholding the punctuality of its delivery promise. While some predictions may be incorrect, the overall benefits far outweigh the occasional inaccuracies.

The healthcare sector provides another excellent example of Predictive AI in action. Hospitals and healthcare providers are using predictive models to anticipate patient admission rates and allocate resources effectively. For instance, Cleveland Clinic used predictive analytics to manage patient flow, ensuring that staff, beds, and equipment were available when & where required, thus reducing wait times by 30%. By forecasting patient demand, hospitals not only shorten the waiting period, but also improve patient care and optimize resource utilization, ultimately leading to better outcomes and cost savings.

There are numerous other examples of Predictive AI delivering on business objectives. Amazon utilizes it for inventory management and enhancing customer personalization. Walmart leverages it to forecast product demand and optimize stock levels to reduce waste and improve customer satisfaction. Shell employs it to predict equipment failures before they occur to ensure smooth operations, minimize downtime, and reduce costs. Climate organizations use it to model climate change scenarios and improve the accuracy of weather forecasting.

I asked ChatGPT for more examples, and this is what it responded with: "A wide array of industries derive significant value from the application of Predictive AI. In the financial services sector, institutions harness sophisticated predictive algorithms for credit risk assessment, fraud detection in transactions, and strategic risk mitigation. By deploying advanced data analytics frameworks to scrutinize temporal transaction histories and client profiles, banks enhance lending decisions, efficiently identify anomalies, and mitigate associated financial exposure. In the education sector, predictive modeling techniques facilitates early identification of at-risk students, thereby enabling bespoke pedagogical interventions. Concurrently, in the agricultural sector, predictive AI systems optimize irrigation strategies, augment agronomic yields, and bolsters food security through tech-enabled smart farming."

Beyond the Buzz: Where Predictive AI Shines

The true promise of AI lies in its ability to enhance processes autonomously, and Predictive AI excels at this by providing critical insights that companies can take action on. GenAI is extremely useful when it offers creative solutions in customer experience, legal research, and coding. However, it is not reliable enough to operate independently in the real-world situations businesses face.

Today, while many organizations recognize the potential value of AI, only a few are managing to scale their AI projects successfully. One of the main challenges they face is the lack of alignment between AI initiatives and business goals, leading to program failures. Thus, for those seeking practical gains, the value is in focusing on specific use-cases where AI, can make a tangible difference.

Hence, despite the excitement around generative models, Predictive AI should be preferred by businesses seeking immediate, measurable operational impact in the near future. Balancing expectations and targeting real-world applications, rather than overhyping the technology, will ensure businesses make progress where it truly matters.

Having said this, the line between GenAI and Predictive AI is slowly blurring. For instance, GPT-4o can already perform complex regression analysis correctly, which is traditionally a strength of Predictive AI. Throughout this article, I have talked about the difference between GenAI and Predictive AI, but it is imperative to state that eventually, there won't be any distinction left. As technology advances, the capabilities of both GenAI and Predictive AI will ultimately converge, offering more hybrid applications that leverage both creativity and precision in real-time. This convergence suggests a future where AI systems are not limited to distinct categories but instead serve multifunctional purposes across various industries.

Conclusion: Let Practicality & Adaptability Guide Your AI Strategy to Future-Proof Your Company

It is crucial that businesses balance the allure of AI with the practical benefits it brings. In the short-term, predictive models are driving real change by enabling companies to act on probabilities and continuously improve. Streamlined supply chains & logistics, more efficient operations, reduced emissions, and optimized marketing are just a few examples of Predictive AI's potential when thoughtfully applied.

Simultaneously, businesses should also build capabilities that allow them to respond flexibly to change. AI is anticipated to contribute $15.7 trillion to the global economy by 2030 , and is projected to create a net gain of 12 million new jobs by next year. These shifts underscore the need to invest in upskilling employees, fostering a culture of experimentation, and staying attuned to evolving AI technologies.

Organizations must develop resilience, as the speed and nature of technological advancements are often uncertain. To thrive in the unpredictable AI-driven landscape, it is essential to adopt a forward-looking approach in your organizational strategies, utilizing any type of AI tool that aligns with your business needs. Focusing on practical applications that deliver tangible outcomes, while being prepared to evolve as new tools and capabilities emerge, will ensure long-term success.

The future is certainly uncertain, but we must be adaptive - it's the only way we can survive!

?

TL;DR: GenAI - Good, Predictive AI - Great, Future - ?

Notes and References

I have provided several links within the article itself, embedded in the images and texts. Here are the other references:

https://www.vellum.ai/blog/100-must-know-ai-facts-and-statistics-for-2024

https://tweet.ps/post/details/artificial-intelligence

https://samosov.ru/artificial-intelligence-ai-definition-examples/

https://brightonapito.com/simplifying-artificial-intelligence/

https://butlerscryptohelp.com/ticklab-revolutionising-finance-with-ai-powered-quant-hedge-fund-and-e-d-i-t-h/

https://www.followlist.com/blog/ai-in-digital-marketing-the-ultimate-guide

https://redresscompliance.com/ai-demand-forecasting/

https://www.techtarget.com/searchenterpriseai/tip/Generative-AI-vs-predictive-AI-Understanding-the-differences

https://dev.to/pavanbelagatti/learn-generative-ai-for-free-e-book-9k1

https://www.slideshare.net/slideshow/understanding-genaillm-and-what-is-google-offering-felix-goh/263757733https://bigthink.com/series/legends/eric-siegel-generative-ai/

https://www.deloitte.com/global/en/services/legal/services/generative-ai-legal-departments.html

https://openai.com/index/codex-apps/

https://cloud.google.com/transform/101-real-world-generative-ai-use-cases-from-industry-leaders

https://appian.com/blog/acp/process-automation/generative-ai-vs-large-language-models

https://bernardmarr.com/generative-predictive-prescriptive-ai-what-they-mean-for-business-applications/

https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-top-trends-in-tech

https://www.mckinsey.com/capabilities/quantumblack/our-insights/ten-red-flags-signaling-your-analytics-program-will-fail#/

https://www.mckinsey.com/featured-insights/artificial-intelligence/global-ai-survey-ai-proves-its-worth-but-few-scale-impact#/

https://www.goldmansachs.com/insights/articles/ai-investment-forecast-to-approach-200-billion-globally-by-2025

https://www.pwc.com/gx/en/issues/analytics/assets/pwc-ai-analysis-sizing-the-prize-report.pdfhttps://www.flexos.work/learn/ai-statistics

https://www.forbes.com/sites/bernardmarr/2024/10/08/the-next-breakthrough-in-artificial-intelligence-how-quantum-ai-will-reshape-our-world/

https://www.forbes.com/sites/jonathanreichental/2023/11/20/quantum-artificial-intelligence-is-closer-than-you-think/

https://chatgpt.com/ - For research

https://otter.ai/home - For transcription

https://app.grammarly.com/ - For proof-reading

Finally, I am not an AI expert, so forgive me for any errors or oversights in this article. This piece is my attempt to understand and contribute to the discussion.