What are the Benefits of AI?

Unless someone has been completely disconnected from current events since before the pandemic, it's safe to assume they've heard of AI. Over the past 18 months since ChatGPT was launched in late 2022, AI has become a widely discussed topic, from everyday conversations to high-level gatherings like the World Economic Forum in Davos. Despite the diverse nature of these discussions and the varying levels of expertise among participants, they all share a common goal: understanding AI, its impacts, and its implications.

There seems to be a prevailing belief—or perhaps a wish—that merely associating something with AI will automatically draw attention to it. While this might have been true in 2023, it no longer holds today. What may not be as widely recognized is the diversity within AI itself, with some types having existed long before ChatGPT.

Moreover, these different forms of AI have distinct implications in terms of the required hardware and software support, as well as their practical applications. As understanding of these nuances deepens, so does the realization that simply invoking the term "AI" is insufficient. The conversation needs to delve into the specific problems being addressed, how AI is being leveraged to solve them, and who stands to benefit.

Conventional versus generative AI

Before exploring the maturing AI ecosystem and the emerging solutions it offers, it's important to take a step back and clarify two primary types of AI: conventional AI and generative AI. Most people's exposure to AI, largely fueled by the hype surrounding ChatGPT, revolves around what is more accurately described as "generative AI". However, there exists another, less recognized but more prevalent form known as "conventional AI."

The key distinction between generative AI and conventional AI lies in their operational characteristics. Generative AI excels in generating novel content based on provided prompts, whereas conventional AI specializes in recognizing predefined patterns and executing actions accordingly.

In essence, while the latter focuses on pattern recognition, the former is geared towards pattern creation. A straightforward example was demonstrated by Jensen Huang at GTC 2024: conventional AI gained prominence with the AlexNet neural network model in 2012, capable of identifying a cat in a picture of a cat. In contrast, generative AI can produce a picture of a cat in response to a text prompt like "cat."

Another differentiating factor is the resource demands for training and inference in each AI type. Generative AI models typically require vast computational resources—akin to a data center's worth of CPUs and GPUs—to be adequately trained. Conversely, conventional AI training can often be accomplished with a single server's resources.

Similarly, for inference tasks, generative AI may necessitate substantial processing power, whether centralized or optimized for edge computing. Traditional AI inference, on the other hand, can be performed with microcontroller-level resources.

While resource requirements vary depending on model complexity and data scale, these comparisons generally hold true for widely used models.

The promise of AI: limitless potential

As the landscape of AI solutions matures, mere mention of AI is no longer sufficient. To establish credibility and competitiveness, a refined strategy, positioning, and demonstration of solutions are essential. Prospective users are no longer satisfied with technology showcases depicting puppies on the beach; they seek tangible value that directly addresses personal or enterprise challenges.

The beauty of the AI ecosystem lies in its diversity, with numerous companies striving to answer these critical questions. At the recent Mobile World Congress (MWC), Qualcomm and IBM stood out for their innovative use of both conventional and generative AI, targeting consumers and enterprises respectively.

Qualcomm showcased real-world applications harnessing AI. Their Snapdragon X80 5G modem-RF platform employs AI to dynamically optimize 5G performance based on user applications and RF conditions, achieving optimal efficiency.

Meanwhile, Qualcomm's generative AI solutions are ushering in a new era of AI-powered smartphones and PCs, empowering users with innovative image and video manipulation capabilities directly on their devices.

IBM, on the other hand, is focused on enterprise solutions through watsonx, featuring a call center assistant that leverages both conventional and generative AI to streamline tasks and optimize resource utilization.

As enterprises integrate AI into their workflows, the need for tailored solutions becomes apparent. IBM's watsonx platform enables enterprises to manage data responsibly, develop custom applications, and ensure governance.

Looking ahead, we are poised to explore the untapped potential of AI. As the global economy embraces AI-driven value creation, the horizon expands with possibilities yet to be imagined.