The Shen Translation Bureau is a translation team under 36Kr, focusing on technology, business, career, life and other fields, and introducing new technologies, new perspectives and new trends from abroad.
Editor's Note: Technically speaking, ChatGPT is not a breakthrough, as the technology it uses is not new. The reason it has become the fastest-growing internet product in history is the chat interface. Revolution does not necessarily come from technological breakthroughs, but sometimes from changes in perception. The article is from the translation team.
The success of ChatGPT has even surprised its creators. The technology itself is not new - it's just the way it's presented that has changed. In her latest article, Rhea Purohit explores how psychological factors, rather than technological capabilities, are the key to driving the widespread adoption of revolutionary technologies. She compares the rapid rise of ChatGPT to the role of the Macintosh in popularizing personal computing, revealing that understanding human psychology may be the key to unlocking the true potential of AI.
——Kate Lee
November 20, 2022, about two years ago, was the day ChatGPT was released.
After its release, the application quickly became a hit.
The whole world was excited, fearful, and even a little skeptical.
But in the first few months, the executives at OpenAI were... a bit puzzled.
Why?
From a purely technical perspective, ChatGPT is not a huge breakthrough in advanced technology. In fact, it's not new. OpenAI's GPT model has been around since 2018, and the original ChatGPT was just a fine-tuned version of GPT-3.5, with most of the technology already available through the API before launch. Yet, despite this, ChatGPT became one of the fastest-growing applications on the internet, with an estimated 100 million monthly active users just two months after its launch.
The executives at OpenAI were puzzled by this.
Jan Leike, who was responsible for OpenAI's alignment team at the time, said in an interview after the launch, "I really hope to better understand what drove all this - what drove this viral spread... because fundamentally the model isn't that much more powerful than what we had before."
Later, Leike answered his own question: "We made it more human-friendly. It talks to you in a conversational way, it's accessible through a chat interface, it tries to be helpful. That's a huge advance, and I think people recognized that."
The reason ChatGPT became a hit is that it packaged the growing potential of AI in a familiar interface - chat. It didn't create new capabilities, but presented existing functions in a different way. ChatGPT redefined our relationship with AI, with the driving force being a shift in our mindset towards large language models, rather than the raw power of the technology itself.
While the technical barriers to AI progress are certainly relevant, the psychological barriers to our adoption of these technologies are equally important. If, for deep-seated human reasons, ordinary individuals decide not to use them, even the most advanced models may fail to deliver on AI's promises. Let's take a closer look at how we psychologically accept new technologies as a culture, and how this will affect the way we develop and use AI, and in turn, our work and lives.
History doesn't repeat, but it rhymes
Revolutions sometimes come from changes in perception. This is not a new concept - twelve years ago, Rory Sutherland, the advertising guru at Ogilvy, expressed a similar idea: "The next revolution may not be a technological revolution at all, but a psychological revolution - a better understanding of people's values, behaviors and choices, which could bring economic value equal to inventing a hovering car or some new electronic gadget."
The hovering car has not yet appeared, but if you stop and think about the devices we use every day, you'll notice that many of them are due to changes in our mindset, not changes in technology.
Take how you're reading this article, for example.
Regardless of whether you're on your mobile phone, laptop, inbox, or a specific website, you're probably interacting with the computer through a graphical user interface (GUI).
GUI is a way of interacting with computers using intuitive visual elements like buttons, icons, and menus. Before GUI, using a PC meant typing long strings of green alphanumeric characters on a black screen. The GUI, combined with the functionality of clicking on visual elements with a mouse, drove the personal computer revolution. The computer credited with popularizing this technology was the prototype of the Macintosh, released in 1984.
But the Macintosh was not the first computer to use a GUI. It wasn't even the second or third (the second was the Xerox Star in 1981, and the third was Apple's Lisa in 1983, both of which had GUIs and pointing devices). The reason the Macintosh occupies a place in history is not because of its technical specifications - the special thing about this device was that it popularized the idea of a "user-friendly" computer. This idea was even reflected in the Macintosh's design, making it look like a symmetrical "face".
The disk drive was moved to the bottom right, making the machine look like a strange long face.
The Macintosh did not bring significant revenue to Apple. It was slow, incompatible with multiple applications, and had laughably small memory. But it led the personal computer revolution because it successfully sold the idea that "anyone can - and more importantly, anyone wants to - use a computer". Before the Macintosh, computers were mostly confined to the back office, used only by a few who knew how to use them. The Macintosh brought them into the living room, making the computer - a complex and intimidating machine - more accessible. It fundamentally changed people's perceptions of existing technology. This is why I believe the next major breakthrough in AI will have nothing to do with algorithms, data, or computing power - but with you. More precisely, with your mindset.
What really drives the adoption of new technologies
Whenever a new large language model is released, my social media feed is flooded with posts measuring the changes in the model's technical capabilities. There are also countless predictions about how more powerful models will affect human survival. We've fallen into the trap of measuring progress solely by technological advancement, always anxiously awaiting the next breakthrough.
AI models may indeed improve over time. However, as the examples of ChatGPT and the Macintosh show, the path from innovation to impact is not smooth. Our mindset plays a crucial role in deciding whether we want to use a product or not.
I was curious to find out what psychological factors influence the adoption of new technologies. After some research on Perplexity, I found the answer in the Unified Theory of Acceptance and Use of Technology (UTAUT). As the name suggests, UTAUT integrates research on how humans integrate new technologies into their work and life. The theory was first proposed in 2003 by a research team including Dr. Viswanath Venkatesh, a business professor at Virginia Tech. Venkatesh studied technology implementation in various settings, initially looking at how employees adopted office technologies, including proprietary accounting software and an online video conferencing platform to reduce the need for face-to-face meetings. Venkatesh later expanded the UTAUT framework to study consumer adoption of new technologies. Over the years, UTAUT has become an important framework for understanding technology adoption in different contexts.
UTAUT explains how humans perceive the adoption of new technologies based on four main variables: performance expectancy, effort expectancy, social influence, and facilitating conditions. Each variable provides a lens for examining the development, promotion, and use of AI products.
Performance Expectancy
When OpenAI launched the o-1 model in September 2024, the company promoted the model through a series of videos demonstrating its exceptional reasoning capabilities. In the videos, it could crack Korean passwords and assist a geneticist in researching rare disease cases. As large language models are versatile and multi-functional tools, these videos showcased the practical applications of the technology to users, thereby shaping their performance expectancy.
Performance expectancy refers to users' perceptions of the usefulness of a new technology for themselves. In professional settings, this is often reflected in improved user efficiency, making it an extrinsic motivator - behavior driven by external rewards or pressures rather than internal satisfaction or enjoyment.
Effort Expectancy
Consider the difference between the mainstream success of AI and the slower adoption of cryptocurrencies. A Reddit user pointed out that cryptocurrencies are "complex, cumbersome, and require a lot of focused knowledge to avoid losses," while the conversational interface of large language models makes AI easy to pick up, positively impacting users' effort expectancy.
Effort expectancy refers to users' perceptions of the ease of use of a product. This variable is particularly important in the early stages of use and gradually diminishes over time and with continued use.
Social Influence
While all large language models can answer your questions, Perplexity aims to keep your curiosity alive. Its CEO frequently posts abstract, thought-provoking images on X, and the minimalist design of Perplexity's merchandise (such as hoodies and hats), as well as the carefully selected sans-serif fonts on its website, all contribute to this atmosphere. The company has built its brand around cultivating curiosity.
Perplexity's brand experience designer, Phi Hoang, explained the origin of the company's brand, stating that they avoid focusing on the specifics of AI models in their information delivery, as end-users are not concerned with these details. In terms of UTAUT, Perplexity scores highly on the social influence dimension, helping it build a loyal niche of advanced users.
Social influence refers to users' perceptions of how others, especially those they respect, view the use of a new technology. It is, to some extent, about whether users perceive the use of the technology as "cool."
Facilitating Conditions
Imagine you're sitting in a cafe with a co-founder, casually coding with the AI code editor Cursor. Suddenly, the app crashes, and you start complaining in frustration - only to have someone come over and fix it for you, who turns out to be a Cursor co-founder.
While this level of customer service may be difficult to scale, Cursor would likely score highly on the facilitating conditions variable.
Facilitating conditions refer to users' perceptions of the availability of a platform to express their concerns and the resources to help them use the product.
A New Perspective on Technological Progress
In the past two years, Jan Leike left OpenAI and started working at Anthropic. He may no longer be puzzled by the viral success of LLMs. The UTAUT framework can explain Leike's questions about why ChatGPT achieved such success, and the academic community has already published some early research exploring how this theory applies to the adoption of AI tools, proposing additional variables such as users' trust in autonomous intelligent systems and their attitudes towards AI risks.
I also see the essence of this theory being applied to new user interaction features in AI. For example, Claude's "Computer Use" functionality allows the model to perform tasks like browsing the web, creating and editing files, or running code, which aligns with effort expectancy for tools. ChatGPT's "Canvas" feature allows users to work in a collaborative workspace using AI, rather than being limited to a chat interface, positively impacting performance expectancy.
As leading AI companies push the boundaries of technology, we, as developers, leaders, or learners, can also create the next AI breakthrough by understanding the psychological factors that drive or hinder user adoption. The future of AI lies not only in better algorithms, data, or computing power, but also in our relationship with the technology.
