Is the AI hype justified? It depends on who you ask. Artificial intelligence (AI) is populating different areas of the Gartner hype cycle. Chatbots? Disillusion. Machine learning? Great, but expectations have been lowered a bit. Responsible AI? Could be the next big thing! Understanding AI is hard because it is not one thing, and the technology has many use cases. In this article, I’ll provide a simpler way to understand AI (a bit): Understand the AI market.
The basics of AI have been promoted widely by consultancy firms like PwC and McKinsey & Company: AI has the potential to add trillions of dollars to the world economy in the next decade, although the impact depends on policymakers and the associated market dynamics. That impact will not impact regions easily, as China (potential 26% boost in GDP by 2030) and the U.S. (14.5% boost in GDP by 2030) will see more gains than other areas, like, for example, the EU (10% boost in GDP in the same time frame).
Specific use cases also see more benefits than others, with the big chunks going to customer-centric topline activities such as customer service, recommender systems, pricing, ads and lead generation. A second high-impact area is value chain optimization, with use cases in predictive maintenance, yield optimization, smarter logistics and inventory optimization.
Break things down further, and you’ll find that AI is mostly a multibillion-dollar software market, be it with lower gross margins than traditional software firms (Andreessen Horowitz, 2020). The use cases for this software per sector are detailed by firms such as DataRobot and C3 AI. This software industry also has a “dark side,” most AI use cases require labels, known as “ground truth.” For this, an entire data labeling industry is set up, reaching $3.5 billion in 2024. One could consider this the blue-collar work of the future.
Saying AI is software alone would miss another big trend affecting the industry from the other side: specialized silicon. Companies like Apple, Google, Nvidia, Graphcore and Hailo are making their own chips specialized for AI applications. Google’s Tensor chip, featured in the new Pixel 6, is a clear example of how AI applications such as language understanding can run much more smoothly on optimized silicon in smartphones.
Players like Graphcore, Nvidia and Google (TPUs) show that optimized hardware in cloud pods can handle heavier and heavier workloads, at lower costs and with greater efficiency. On the other end, the Israeli firm Hailo focuses on edge computing, technically allowing for local processing of signals, limiting the need for cloud compute.
A final consideration to understand AI today is market regulation. Over 50 countries either have designed or plan to design their own AI strategies in recent years. AI was mentioned eight times in Congress in the year 2013-2014 and 486 times in 2019-2020. Not surprisingly, countries today are implementing regulations (e.g., GDPR, data pooling laws), industry protection, government stimuli and incentives for top talent. These factors will influence how the AI industry will unfold in the coming years.
A more in-depth webinar on this topic is available here.