The “hype cycle” is about the hype—not the technology

Insight categories: AI and MLPerspectiveTechnology

The hype cycle has little to do with the merits of a particular technology. It simply has to do with the amount of publicity the technology has received. In particular, if the publicity jumps ahead of what the technology can immediately deliver, then the technology quickly gets labeled as “over hyped”. This is not the ‘fault’ of the technology—just of the overinflated expectations for immediate benefits that grow up around it.

A case in point is, believe it or not, the world-wide web. Back in 1994, my team set up NeXT Software’s (now Apple’s) first website. At the time, there were only something like 10,000 websites on the entire internet (at this writing there are well over a billion). Even at its beginnings, though, it seemed obvious to me—and to a lot of other people—that Web technology was transformational. However, in the late 1990’s, believe it or not, the Web was considered over-hyped.

With the benefit of 25 years of hindsight it seems almost incredible to us that the world-wide web and the internet could possibly be considered overhyped. If there’s a single technology that truly transformed the world, I think most of us would agree that it’s the Web (plus the internet and the ‘personal’ computer, but those are stories for perhaps another day). The Web and the follow-on technologies it spawned have completely transformed our world, and their impact continues to fill our working and personal lives. Web-related and web-motivated technologies include social media, the cloud, smart handheld devices (phones, tablets, etc.), massive multi-player games, on-line dating, dynamic content creation, shopping and connected cars, and many others. In fact, it’s hard to imagine modern life without the Web, the internet, and its various downstream impacts. We simply take for granted instant access to information, ubiquitous connectivity, pervasive communication, remote device monitoring and control, media when and where we want it, as many others. These are now simply built into the fabric of our lives.

Yet the people who claimed the Web was overhyped in the late 1990’s had a point. At that time, connectivity was limited, and complex graphically rich page renderings were slow. Even when user interactivity was introduced, it was—at first—very simple by today’s standards; essentially form-based. E-commerce emerged very early—within two years of the first static website I mentioned—but issues like payment security were still being worked out and trust was low by today’s standards. And indeed, the naysayers were right in one sense: there was a “dot-com bubble” that burst and struck down many web- and internet-centric companies in the early 2000’s. While this downturn had many causes, one of them was that the “hype” had indeed gotten ahead of the technology.

Why do I bring up this ancient history? I think we’re going to see something similar happen to GenAI, probably this year (2024). Like many people, I am confident that GenAI and the downstream technologies it inspires will utterly transform the world—on the scale that the internet, the world-wide web and their follow-on technologies have done, if not more. Bill Gates is quoted as saying that in the short-run GenAI is overhyped, but in the long run it is under-hyped. I don’t know if Mr. Gates was thinking of the history of the Web when he said this, but I’m sure the analogy must have been on his mind. His remark is an excellent description both of the Web‘s historical adoption curve, and sums up very neatly what I think is likely to come with GenAI.

Today’s tools and technologies make it easy to create a very compelling demo with GenAI.  Today in early 2024, eighteen months after ChatGPT went public in the fall of 2022, many of us, myself included, continue to be stunned by what this technology can do. We are even more excited by what it promises for the future. However, as the POCs move into enterprise-scale deployments and business-critical applications, the problems and gaps will predictably start to surface.

People will realize that data is harder to gather, prepare, curate and keep relevant than they suppose. Approaches that only a few months ago defined the state of the art for GenAI development will change as new approaches are invented—obsolescing systems already built. We’ve seen this already: the “RAG” model (“Retrieval Augmented Generation”) that six months ago was so cool is now being termed the “naive RAG model” and has been replaced by the “advanced RAG model”. Probably, in the new future, it will itself be replaced by other approaches that are even better. Lots of work that was done to work around the 4k token window size supported by popular LLMs has become unnecessary because those token windows are expanding to 128k and are growing larger. People are starting to realize that the GPUs needed to power many GenAI systems are expensive and hard to come by, both physically and even on the cloud. New security vulnerabilities and threats will be discovered and invented. And, of course, hallucinations, bias, and inconsistent answers will plague suppliers and applications.

I think it’s pretty much inevitable that there will be a media (social and other) backlash against GenAI in the near future, and that the technology will be labeled as “over-hyped”. I sincerely hope it does not cause the Armageddon among startups that the “dot-com bust” of the early 2000’s did, but some companies will certainly fall victim to the hype cycle plummeting into what Gartner calls the “Trough of Disillusionment” [].

To reframe a famous phrase in a totally different context, though, my experience of the dot-com era tells me that “the end of the peak hype cycle is the beginning of wisdom”. I think it’s a healthy thing for us all to realize that this technology will not, overnight, transform the world. Like all new technologies, GenAI has rough edges that need to be smoothed out, limitations that need to be discovered and overcome, security and other holes that need to be plugged, and infrastructure that has to be built around it before it becomes commonplace. I also believe that this will happen, and that GenAI and its downstream technologies will fulfill the promise that many of us see in it—and probably faster than we think. The important thing, as technologists, is to realize that the “hype cycle” is simply about the hype—it’s not about the technology. Let’s hope our bosses with the money understand the same thing!


A diverse group of software engineers working collaboratively in a brightly lit office on the design and engineering of an intelligent next-generation AI platform.


Dr. Jim Walsh


View all Articles

Trending Insights

If You Build Products, You Should Be Using Digital Twins

If You Build Products, You Should Be Using...

Digital TransformationTesting and QAManufacturing and Industrial
Empowering Teams with Agile Product-Oriented Delivery, Step By Step

Empowering Teams with Agile Product-Oriented Delivery, Step By...

AgileProject ManagementAutomotiveCommunicationsConsumer and RetailMedia

Top Authors

Ravikrishna Yallapragada

Ravikrishna Yallapragada

AVP, Engineering

Lavanya Mandavilli

Lavanya Mandavilli

Principal Technical Writer

Oleksandr Fedirko

Oleksandr Fedirko

Senior Solution Architect

Mark Norkin

Mark Norkin

Consultant, Engineering

Deborah Kennedy

Deborah Kennedy

Associate Vice President,Client Engagement

All Categories

  • URL copied!