Is the generative AI revolution destined to stagnate? Deep learning skeptic Gary Marcus recently suggested this in his blog, stating that the generative AI “bubble has begun to burst.” Generative AI encompasses systems that create new content—such as text, images, code, or audio—by analyzing extensive datasets. Recent news and analyst reports have raised doubts about the immediate utility and economic value of generative AI, particularly regarding large language models (LLMs).
Historically, skepticism toward new technologies is common. For instance, Newsweek published an article in 1995 declaring the Internet a failure, arguing it was overhyped and impractical. Fast forward to today, and we live in a world transformed by the Internet. This context prompts us to question whether today's skepticism about generative AI might similarly be shortsighted. Are we undervaluing AI's long-term potential because we are fixated on its current challenges?
For example, Goldman Sachs recently questioned the practicality of generative AI in their report titled, “Gen AI: Too much spend, too little benefit?” Additionally, a survey by Upwork revealed that nearly half (47%) of employees using AI struggle to meet their employers' productivity expectations, while 77% reported that these tools have decreased their productivity.
A year ago, Gartner placed generative AI at the “peak of inflated expectations.” However, their latest assessment indicates a shift toward the “trough of disillusionment,” a phase when interest diminishes due to unmet expectations during experimental implementations.
This cyclical pattern in technology adoption isn't new. The buildup of hype around new innovations is a natural human tendency, often overshadowing the hard work required to translate potential into real-world benefits. As analyst Benedict Evans noted, “what happens when the utopian dreams of AI maximalism meet the messy reality of consumer behavior and enterprise IT budgets: It takes longer than you think, and it’s complicated.” Overestimating new systems' promises is a hallmark of bubbles.
This observation echoes Roy Amara's insight from 1973: “we tend to overestimate the impact of a new technology in the short run, but we underestimate it in the long run.” This principle, known as “Amara’s Law,” is particularly relevant in the context of generative AI.
Technology and its supporting ecosystem often require time to mature. Ken Olsen, then CEO of Digital Equipment Corporation, famously claimed in 1977 that “there is no reason anyone would want a computer in their home.” This assertion proved wrong as personal computers eventually became ubiquitous. The development of AI technology will similarly take time.
Considering AI's potential evolution, a 2018 study by PwC identified three overlapping cycles of automation through AI extending into the 2030s: the algorithm wave (early 2020s), the augmentation wave (late 2020s), and the autonomy wave (mid-2030s). Currently, discussions focus on how AI can augment human abilities and work. For instance, IBM states that AI's purpose is to enhance human intelligence, and JPMorgan CEO Jamie Dimon believes AI can “augment virtually every job.”
Real-world applications of AI already demonstrate its ability to enhance human capabilities. In healthcare, AI diagnostic tools improve disease detection accuracy. In finance, AI algorithms bolster risk management and fraud detection. Customer service has also benefited from advanced chatbots that provide 24/7 support, streamlining interactions. These examples reveal that while AI is in its nascent stage, it's steadily improving efficiency across industries.
While augmentation is not the same as full automation and is unlikely to eliminate many jobs, it aligns with the current trajectory of AI development. Like word processors or spreadsheets, once mastered, generative AI can significantly boost productivity, though it won't fundamentally change the world.
Disillusionment stems from the gap between expectations of generative AI as a revolutionary technology and its present reality. Conversations should focus on a realistic timeframe for progress. As venture capitalist Marc Andreessen noted, every failed Dotcom idea could succeed today simply because it requires time.
AI development will inevitably progress, varying across industries and professions. Some areas will experience rapid advancements, while others may lag. This uneven development reflects the current state of generative AI as it navigates its growing pains.
Ultimately, generative AI is poised to be revolutionary, albeit possibly later than some optimistic predictions suggest. The most significant advancements are likely to emerge in the next ten years, coinciding with PwC's predicted autonomy wave. This phase will see AI capable of analyzing data, making decisions, and taking actions with minimal human intervention, leading to applications that seem like science fiction today, such as fully autonomous vehicles.
As AI continues to augment human capabilities, the revolution is unfolding, though perhaps more gradually than anticipated. Perceived slow progress may generate stories of AI falling short, fueling pessimism about its future. However, in line with Amara’s law, AI is expected to mature and fulfill its revolutionary potential.
Gary Grossman is EVP of Technology Practice at Edelman.
Ranking
