Why the generative AI hype is good

Elon Musk predicts artificial general intelligence (AGI) by 2025. Former OpenAI employee Leopold Aschenbrenner says AGI by 2027. Futurist Ray Kurzweil bets on 2029 for AI surpassing human intelligence. OpenAI cofounder Ilya Sutskever declared that language models are “slightly conscious.”

These claims falsely suggest that any limitations or congenital defects will soon be solved, and they might, but not by simply waiting. The tendency to imagine that all we need is more time is dangerous, particularly for business, because business is not a waiting room.

[ This article is an excerpt from Generative Artificial Intelligence Revealed, by Rich Heimann and Clayton Pummill. Download your free ebook copy at the book’s website. ]

A former Google engineer, Blake Lemoine, made headlines in 2022 when he publicly claimed that Google’s chatbot, known as LaMDA (Language Model for Dialogue Applications), had achieved sentience. Lemoine argued that LaMDA exhibited self-awareness and emotions, describing its responses as a “sweet kid who just wants to help the world be a better place for all of us.” Some ideas are so silly that only AI researchers can believe them.

We do not endorse the unsupported claims made by Musk, Aschenbrenner, Kurzweil, Sutskever, and Lemoine. However, we will argue that hype is necessary and beneficial, although it is insufficient for innovation. What we need for innovation is not just hype, but also “getting our hands dirty” through experimentation.

Hype may seem excessive, and at times it is, but it is crucial for innovation as it mobilizes capital, attracts talent, and captures public interest. However distasteful, Amara’s law—that we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run—highlights the role of hype as both a catalyst and a signal for future opportunities. Businesses must balance short-term caution with long-term positioning to leverage these opportunities effectively. In other words, innovation requires navigating the tricky landscape of hype.

Overcoming uncertainty

Critics argue that hype leads to waste and false promises, and this argument certainly has gained momentum with respect to AI. Goldman Sachs published an article in June 2024 titled “Gen AI: too much spend, too little benefit?” Sequoia Capital published an article (also June 2024) titled “AI’s $600B Question.” The article asks, “Where is all the revenue?” In September 2023, the same author published “AI’s $200B Question.” However, the accumulation of knowledge and innovation does not respond to repeated questioning. Innovation does not occur in straight lines or on paved roads, and the absence of immediate returns is not evidence of failure.

We understand the impulse and desire in almost all of us to believe that we wouldn’t need a “costly filter on the back side” if not for hype. The reality, however, is that without hype, we wouldn’t need filtering because no one would care. While the cycle of hype and correction may seem excessive and the exaggerated claims nauseating, the market would stagnate without them due to a lack of interest. The perspective that we don’t need hype overlooks hype’s positive impact in driving interest and investment. Without hype, we would experience a general malaise due to a lack of engagement and enthusiasm.

Venture capitalists anticipate that over half of their investments will fail and eagerly embrace hype as a signal for future opportunities and the tailwinds it brings. Market-based filtering mirrors John Wanamaker’s famous advertising lament: “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” A parallel quote that reflects this reality might be: “Half the money invested in your AI strategy will be lost; the trouble is not knowing which half.” This captures the uncertain nature of investments, where the outcomes remain inherently unpredictable despite due diligence, strategic planning, risk management, and the benefit of hype as a market signal and source of tailwinds.

Investing in general-purpose technologies like generative AI involves even more uncertainty than typical innovation. Generative technologies have broad applications, and finding successful use cases requires knowledge acquired over a longer period. There’s an adage in problem-solving that states 90% of the work takes 90% of the time, and the remaining 10% takes the other 90%. The statement highlights a common issue in project management and technology development: the difficulty of estimating the time required to solve a problem before finishing it. The first 90% often includes planning and explicit knowledge, which is predictable and where progress is made relatively smoothly.

The remaining 10% of a problem typically demands the most effort because it involves the initial limitations of technology and aspects of a problem that we understand the least. This phase involves unforeseen complexities and unexpected challenges that are more time-consuming by definition. We fail to recognize these limitations and challenges until we face them directly, making this phase particularly demanding. At the outset, there is usually a sense of optimism and confidence that a task is straightforward. The initial confidence gives way to the realization that the most challenging aspects of a problem lie in the details we didn’t foresee.

The August 31, 1955 proposal for the Dartmouth Summer Research Project on Artificial Intelligence infamously noted, “We think that a significant advance can be made in one or more of these problems if a carefully selected group of scientists work on it together for a summer [emphasis added].” Such optimism shows that the crucial issues are never apparent during the planning phase. These could be technical challenges, integration issues, or changes in requirements. Addressing these is time-consuming and requires more effort than anticipated. Ultimately, the complexity of problems is a filtering function acquired after getting one’s hands dirty, which takes time to develop.

The role of skepticism

Hype creates new expectations, which is generally good. However, not all these expectations are realistic. Skepticism highlights unrealistic and unmet expectations, which is also good. Yet, skepticism also shapes unrealistic timelines created by people with clean hands. The most costly filtering function imaginable is the one we make on the front side of innovation to prevent problem solvers from getting their hands dirty. We would still be building bridges with woodworking techniques if we had a filter on the front side of innovation. Ultimately, problem-solving and markets are far more effective filtering functions than the hype police.

The attitude that we wouldn’t require filters if not for hype is what we call the hidden challenges of wishes. Some people wish for hype to disappear or implement costly filters on the front end. Others wish for investors to spend money in ways they see fit. While these wishes might seem straightforward and beneficial, fulfilling them can lead to unforeseen complications and unintended effects. In other words, our desires have complex realities, so evaluating wishes with a balanced perspective is crucial.

Regardless, the idea that we don’t need “costly” filters is wrong. Cost on the back side is essential. Without cost pressure, there would be no motivation to improve technology. The cost of failure through competitive disadvantage, market share erosion, or regulation catalyzes accountability, sustainability, and efficiency, which are essential for ongoing progress. Just as no one would care about future technology without hype, no one would care about improving technology without cost.

As Carlota Perez put it, “Nothing important happens without crashes.” Boom and bust cycles are inevitable and necessary to redistribute capital, eliminate unsustainable ventures, and foster long-term, sustainable development. Historically, markets have seen similar patterns, from the “tronics” boom in the 1960s to the dot-com bubble of the late 1990s to the numerous AI winters and summers. Each cycle follows the same initial exuberance, correction, and eventual stabilization trajectory. The interplay between hype and skepticism, driven by the need for corrective cycles, is fundamental to technological progress. Embracing this dynamic allows businesses to navigate the challenges and opportunities of disruptive innovations.

Ultimately, the relationship between hype and skepticism is essential. Hype elevates expectations and imagines a bold and positive future because it is not afraid of the future. The problem with unfettered hype is that it may never find revenue or acquire a customer. Skepticism purges hype that can cloud collective judgment and produce unmoderated enthusiasm but simultaneously creates unrealistic timelines. The skeptic’s role is crucial in maintaining an equilibrium between innovation and practical reality, providing stability in the market.

Amara’s law underscores that hype is a leading indicator that deserves your attention, but not all of it because hype speeds ahead of skepticism and may ignore revenue reflected in goofy goals such as AGI, Turing Tests, and machine consciousness. Amara’s law also emphasizes skepticism as a trailing indicator that deserves your attention, but not all of it because skepticism is sometimes too smart and its hands too clean for its own good. A delicate blend of both begins by understanding the value of hype but never forgetting the importance of skepticism.

We don’t know what problems language models will solve in the future. While we are seriously doubtful of the vacuous claims of AGI, we are confident that humans will find creative ways to apply generative technology, not because we believe the technology has any intelligence but because we are convinced in humanity’s creativity and intelligence to solve new problems natively. Like it or not, hype plays an important part in the journey.

Rich Heimann is a leader in machine learning and AI whose former titles include Chief AI Officer, Chief Data Scientist and Technical Fellow, and adjunct professor. He is the author of Doing AI: A Business-Centric Examination of AI Culture, Goals, and Values and co-author of Social Media Mining using R.

Clayton Pummill is a licensed attorney specializing in complex machine learning, data privacy, and cybersecurity initiatives while building enterprise solutions and support practices for organizations facing machine learning regulations. Active in the technology startup space, he has developed patented technology, co-founded organizations, and brought them through to successful exits.

—

Generative AI Insights provides a venue for technology leaders—including vendors and other outside contributors—to explore and discuss the challenges and opportunities of generative artificial intelligence. The selection is wide-ranging, from technology deep dives to case studies to expert opinion, but also subjective, based on our judgment of which topics and treatments will best serve InfoWorld’s technically sophisticated audience. InfoWorld does not accept marketing collateral for publication and reserves the right to edit all contributed content. Contact doug_dineley@foundryco.com.