Doubt Creeps into the Discussion Over Generative AI: Will the Investment Pay Off?

Doubt is creeping into discussions over generative AI, as industry analysts begin to publicly question whether the huge investments in this technology will ever pay off. The absence of a “killer app” beyond coding co-pilots and chatbots is the most pressing concern, according to critics in a recent Goldman Sachs Research letter. Additionally, challenges related to data availability, chip shortages, and power consumption contribute to headwinds faced by the generative AI sector. However, many remain optimistic about the long-term prospects of GenAI for both business and society.

The Historical Context of Tech Hype

The sheer amount of hype surrounding generative AI over the past year and a half has garnered the attention of seasoned tech journalists, particularly those who lived through the dot-com boom and bust at the turn of the century. This historical perspective includes the rise of cloud computing and smartphones, notably with the introduction of Amazon Web Services and the Apple iPhone in 2006 and 2007, respectively.

Following this, the big data boom of the early 2010s became a new obsession, culminating with the coronation of Hadoop as The New New Thing. After Hadoop’s eventual collapse, the focus subtly shifted to AI, and there have been various other technological innovations vying for attention along the way, including Blockchain, 5G, and self-driving cars. However, none of these technologies gained significant traction, leading to skepticism about their actual value.

Is Generative AI a Fad or the Future?

Since OpenAI launched its large language model (LLM), ChatGPT, in late 2022, the level of hype surrounding generative AI has reached unprecedented heights. Yet, some analysts, including those at Goldman Sachs, are beginning to draw parallels between generative AI and other technologies that ultimately failed to deliver on their promises.

In a much-discussed report from June titled “Gen AI: too much spend, too little benefit?”, editor Allison Nathan raises critical questions about the viability of generative AI in the near term. She points out that tech giants and various companies plan to invest approximately $1 trillion in capital expenditures to enhance their AI infrastructure, yet she notes that this spending has not yet produced significant benefits beyond minor efficiency gains among developers.

The Challenges Ahead: Automation and Resource Constraints

Prominent MIT Professor Daron Acemoglu argues that only a fraction of the tasks that AI is intended to automate will actually be automated in a cost-effective manner. He estimates that merely 5% of all tasks will be automated within the next decade, projecting minimal productivity growth as a result. While Acemoglu acknowledges the potential of Generative AI to revolutionize sectors like scientific research and development, he warns that these transformative changes are unlikely to materialize quickly.

Moreover, the prediction that increasing the quality and quantity of data will accelerate progress in generative AI may be overly simplistic. Acemoglu emphasizes that simply increasing data input—such as adding more Reddit content to a model—does not necessarily enhance its practical utility in real-world applications like customer service.

Power Demands and Market Realities

Another significant challenge lies in the supply constraints of semiconductors necessary for training generative AI models. Nvidia has greatly benefited from this demand, experiencing a revenue surge of over 260%, which helped elevate its market capitalization to over $3 trillion, making it one of the most valuable companies globally.

Goldman’s Jim Covello comments on the formidable barriers facing startups attempting to rival Nvidia’s dominance, noting that many have tried and failed over the past decade. The immense costs associated with training and deploying generative AI further complicate the possibility of achieving seamless productivity gains.

The Mixed Perspectives on Generative AI\’s Future

Despite the prevailing skepticism, not everyone shares a bleak outlook on the future of generative AI. Goldman’s senior global economist, Joseph Briggs, asserts that GenAI could potentially automate 25% of all work tasks, leading to a significant increase in U.S. productivity over the next decade. He posits that while GenAI might initially replace existing tasks, it will also generate new job opportunities, thus fostering economic growth.

Similarly, analyst Kash Rangan expresses confidence in the infrastructure buildout surrounding generative AI, suggesting that we are merely in the early stages of what could evolve into its killer app. He likens this progression to historical computing cycles, where infrastructure development precedes the emergence of impactful applications.

Ultimately, while the promise of generative AI remains tantalizing, its success hinges on overcoming significant challenges. Whether or not generative AI will yield returns before time runs out is a question that demands careful scrutiny as the landscape continues to evolve.