According to the Gartner hype cycle , the least amount of time a product takes to hit the ‘plateau’ of expectations is two years. Not with ChatGPT, though. The hot chatbot has shattered all records of a product lifecycle, going through all stages of the cycle within 3 months.
Launched in November-end last year, ChatGPT has already been through the innovation trigger, inflated expectations, disillusionment, enlightenment, and is now reaching a mature period of measured expectations, leading to industry adoption. A contributing factor to this might be the bot’s meteoric growth, which scaled to 10 million users within 40 days. For contrast, Instagram took almost a year to reach the same milestone.
The hype cycle is a concept created by research firm Gartner to depict the relationship between time, the maturity of technologies, and expectations associated with it. The cycle generally progresses through five phases before reaching maturity. Let’s delve deeper into these phases and how ChatGPT raced through each of them.
As the name implies, this phase is birthed out of years of research and technology. The technology trigger phase is characterised by the publishing of proof-of-concepts, research papers, and the beginning of mainstream media coverage. In the case of ChatGPT, OpenAI’s research into LLMs like GPT-2, GPT-3, and GPT-3.5 were the triggers that led to the creation of ChatGPT, laying the groundwork for the chatbot.
OpenAI had been working on these algorithms and others for years before the launch of ChatGPT. While the company chose to not release its models to open access, and instead went the API way, its steady pace of innovation cemented its place in the minds of AI enthusiasts. This could have been a ‘sleeper’ trigger, where expectations for OpenAI’s products slowly increased over the years, compounded by the mystery surrounding their closed nature. Once this trigger was pulled, ChatGPT rode the wave of hype all the way to the peak.
Peak of Inflated Expectations
This phase represents the peak of the hype cycle. In this stage, media coverage and user expectations reach an all-time high, representing the peak of hype surrounding the product. ChatGPT reached 1 million users within just 5 days of launch, with media attention further amplifying the expectations associated with the service. We can place this around December 11-17, 2022, as evidenced by this spike in the Google Trends graph.
As more people began to search forChatGPT, the conversation also trended around bold predictions regarding ChatGPT’s capabilities for disruption. Articles with headlines such as ‘ChatGPT Could Transform Academia’ or ‘ChatGPT May Doom High School English Classes’, painted a embellished picture of what the bot was capable of doing — a reliable portent of inflated expectations which would come crashing down.
Trough of Disillusionment
The trough of disillusionment is a phase wherein experimentation of the technology begins to shed light on its shortcomings. As more people began to use ChatGPT, the cracks began to show, as the bot’s token responses began to fray the nerves of users. Further, reports emerged of the rampant disinformation spread by ChatGPT, especially in political contexts.
This experimentation also exposed the various exploits present in the bot. Users found what they termed ‘jailbreaks’, which were ways to get the LLM to sidestep its programming and provide responses that violate OpenAI’s content policy. This, along with other basic mistakes by the chatbot — like a failure to multiply -1x-1x-1 or the inability to write a sentence ending in a specific letter — led to detractors of the bot growing in number. However, the true capabilities of the bot were not forgotten, resulting in a slow spike in adoption by users who found value in the service.
Slope of Enlightenment
This phase is categorised by the world achieving a larger understanding of what the product is capable of as a whole. The second- and third-order effects become clear, and more parties aim to contribute to the ecosystem and solve some of its more obvious issues.
After the initial wave of disappointment that ChatGPT was not omnipotent wore off, an ecosystem began to spring up around the bot. Many community members added extensions to ChatGPT, making it more capable and powerful, such as this extension to make it a Grammarly substitute, or this one which added the chatbot’s capabilities to Google. This phase also saw the rise of tutorials that provided insights into how to get the most out of ChatGPT, further increasing its usability and utility.
Plateau of Productivity
This phase is the ideal position for all products on the hype cycle, as it represents market maturity and widespread adoption. Many technologies don’t reach this point, as evidenced by the metric of ‘obsolete before plateau’ included in Gartner’s visualisations. However, it seems that ChatGPT has traversed the hype cycle and is currently on its journey to the plateau.
Not only did ChatGPT complete all stages of the hype cycle in record time, spurred by a strong network effect and fascination towards AI, it also paved the way for industry adoption — the true sign of a mature product. Google and Microsoft have both decided to integrate ChatGPT-like chatbots into their search functionalities, and the industry as a whole has moved past the shortcomings of LLMs and is looking to adapt to the coming wave of AI-generated content.
In this respect, ChatGPT has not only bypassed the traditional timeline associated with the hype cycle, it has also gone beyond it and become a generational leap in technology. It holds the potential to change the way people interact with the web, even when adopted slowly and with a measured approach. Moreover, the media attention to ChatGPT’s meteoric growth might also bring the focus on the problems plaguing LLMs as a whole, leading to the creation of more powerful AIs with strong moral and ethical guidelines.