How AI Got a Reality Check

In partnership with

The gold standard of business news

Morning Brew is transforming the way working professionals consume business news.

They skip the jargon and lengthy stories, and instead serve up the news impacting your life and career with a hint of wit and humor. This way, you’ll actually enjoy reading the news—and the information sticks.

Best part? Morning Brew’s newsletter is completely free. Sign up in just 10 seconds and if you realize that you prefer long, dense, and boring business news—you can always go back to it.

Join 4.3 Million Readers Now

In recent years, artificial intelligence (AI) has captured global attention, driven by revolutionary breakthroughs that have both amazed and raised critical questions about its future. ChatGPT, a large language model developed by OpenAI, is one of the most prominent examples. Within days of its release, it became a sensation, showcasing its ability to generate human-like responses, compose poetry, and even admit its own mistakes. But beyond its viral success lies a story of soaring ambitions, mounting challenges, and an industry grappling with the limits of its own innovations.

The Rise of AI

The roots of AI stretch back to the 1950s, with visionaries like Alan Turing laying the foundation. However, it wasn’t until recent years that AI experienced a rapid acceleration, thanks to advancements in computing power, access to large datasets, and the rise of models like ChatGPT. These tools sparked a wave of interest in generative AI, revolutionizing industries and inspiring startups to integrate these technologies into their operations.

But as the technology advanced, so did expectations. AI's evolution depended on "Large Language Models" (LLMs)—massive systems trained on vast amounts of data sourced from the internet. These models could generate eerily human-like text and perform increasingly complex tasks, seemingly heralding a new era of intelligent systems.

The Limits of Large Language Models

Despite their success, LLMs are now reaching a plateau. Scaling these models has become prohibitively expensive. Training a cutting-edge AI model, as Anthropic (a leading AI lab) notes, can cost upwards of $100 million, with future models potentially reaching $100 billion.

Moreover, the "low-hanging fruit" of easy gains is gone. The internet, the primary source of training data, has been largely scraped, forcing companies to turn to curated datasets created by experts. These specialized datasets come at a high cost, often requiring input from PhD-level researchers or industry specialists. Some firms have even begun experimenting with "synthetic data," where AI-generated content trains new models. However, this approach is still unproven and fraught with challenges.

Cost Pressures and Investor Concerns

The financial strain on AI companies is immense. While OpenAI has secured paying business customers, the revenue generated often pales in comparison to the billions spent on research and development. For now, the industry relies heavily on investor funding. But as the costs rise and the return on investment becomes murkier, questions loom about the long-term sustainability of this spending spree.

Despite these challenges, companies are doubling down, driven by the promise of groundbreaking advancements. Recent developments, such as reasoning-based AI models that “think” through problems more effectively, hint at what’s possible. Another area of focus is the emergence of "AI agents"—models capable of not just generating text but completing complex tasks like booking travel or integrating code into software systems.

The Road to AGI

Perhaps the most tantalizing and controversial goal of AI research is Artificial General Intelligence (AGI)—an AI capable of thinking, reasoning, and solving problems across disciplines, much like a human. While AGI could revolutionize the world, it also raises ethical, social, and existential concerns. Will it surpass human capabilities? Could it eventually work independently of us—or even for itself?

Predictions for AGI vary wildly. Some believe it could emerge within a decade, while others argue it might take a century—or never happen at all. The challenges faced by today's AI systems suggest that the road to AGI may be far more complex than initially imagined.

Conclusion

AI has already transformed the world in ways few could have anticipated. Yet, as we celebrate its achievements, we must also confront its limitations and the immense resources required to sustain progress. The journey ahead will be marked by innovation, setbacks, and difficult decisions about how to balance the pursuit of intelligence with ethical and financial realities.

One thing is certain: AI's story is far from over, and its next chapter will be one to watch.