The AI Burnout Is Real
We were sold magic, but the magic is wearing off
Canva has unveiled their “ambitious” new product features called “Magic.” And at this point, I can’t help feeling like there’d be more fanfare if a tech company announced their commitment to not shoehorning a grab bag of chatGPT-inspired LLM features into every piece of software on the internet.
Even when these capabilities aren’t vapourware, there is a growing sense of exhaustion. When everything is pitched to us as “magic”, and none feels magical at all, the inevitable outcome is AI burnout. Yes, I can now ask Canva to regurgitate Shakespeare back in a bespoke sonnet about eCommerce and then shove it into an Instagram template. But I can do the same thing in every other app, in the same flow, with the same outcome. And when the fanfare that precedes each of these feature releases is magnified to an almost cartoonish degree, it’s hard to come up with a reaction beyond…well, a yawn.
Much of the excitement around AI comes from unrealistic expectations. We anticipate human-like intelligence and self-awareness from AI. There is an assumption that AI can and will solve all of humanity’s problems, making decisions for us without any errors or biases. Some believe AI will rapidly advance to superhuman levels, like sci-fi visions of the future. And these expectations are perfectly understandable: it’s precisely how AI tools are being positioned by their creators.
But so far, the life-changing revolutions remain mirages on the horizon. The actual capabilities of current AI systems are far more limited, flawed and biased than the hype suggests. Most real-world “AI” consists of LLMs – Large Language Models. These operate within narrow parameters, not as a promised artificial general intelligence; for all the potential of AI, today’s systems need more flexibility, nuance and context.
AI lacks flexibility and adaptability. While AI can excel at performing specific tasks, it needs help adapting to new or unexpected situations, primarily because AI is based on pre-programmed rules and algorithms and can only operate within the parameters it has been given. This lack of flexibility can be frustrating for users who expect AI to be able to handle a wide range of tasks and situations.
Magic-chasing users will quickly encounter AI’s inability to understand context and nuance. AI operates on data and algorithms, which means it needs help understanding the subtleties of human language and behaviour. FAI-powered chatbots struggle to understand and/or express human elements of sarcasm or irony, leading to user misunderstandings and frustration. This limitation can be particularly problematic when human interaction is essential, as in customer service or healthcare.
AI struggles with bias and fairness, with profound ethical implications that are an unpleasant reality for users expecting digital utopia. Because AI is based on data, it can be influenced by the prejudices and assumptions of the people who create it. If an AI algorithm is trained on biased data against a particular group of people, it will perpetuate that bias in its decision-making. And as much as that may feel like an accurate reflection of JK Rowling’s worldview, I can’t say it feels particularly magical.
AI promises to make creativity effortless but demonstrates an absolute lack of creativity and intuition. While AI can perform tasks that require high precision and accuracy, it struggles with jobs that require creativity or intuition. An AI-powered artist may be able to replicate existing styles and techniques, but it is unlikely to produce genuinely original or groundbreaking work without significant effort and investment from the creator. This limitation can be frustrating for users who expect AI to be able to innovate and create in the same way that humans can.
Finally, the LLM flavour of AI has an innate – and misrepresented – inability to replicate the complexity and nuance of human emotions and relationships. While AI can be programmed to recognise and respond to specific emotions, it lacks the depth and complexity of human emotions and relationships.
When users primed to expect AI tools to be the second coming of Christ come up against perfectly reasonable LLM limitations, the result is disillusionment with AI as a concept.
And we’re left with an emptiness that brings a renewed interest in the earlier days of the internet, before algorithms and personalisation. There is nostalgia for the quirky online communities that sprouted organically, free from corporate influence. People yearn to reduce their screen time and escape the endless personalised feeds of social media. Seeking more authentic human connections online has become appealing after so much time spent alone with machines. Many wish to get away from the hyper-personalization of content by AI systems. They want to reclaim the serendipity and surprise from varied human recommendations. This has sparked a push towards online communities built on human curation, not AI. People appreciate the messy, unfiltered side of the internet that reflects our shared humanity. The rejection of the AI mirage rekindles excitement about the organic growth of the internet.
At its best, the internet can connect us across differences and distil shared truths from many perspectives. Diversity of opinions, backgrounds and contexts is its strength. The complexity and nuance of human lives resists being reduced to data points. There is a craving for the creativity and unpredictability that only humanity provides. While AI has proven capabilities in narrow tasks, it lacks the imagination to leap beyond its programming. The serendipity of random human encounters online creates joy and meaning. Hyper-personalized AI systems cannot replicate that, particularly when hyper-personalization is more wishful thinking than practical product features.
I don’t think I have a viscerally adverse reaction to AI. Even as a writer, I don’t feel particularly endangered by the newfound ability of the entire population of the internet to spit out 2,000 words of median content at a moment’s notice, and I do appreciate the AI-powered editing and functionality of my toolset.
But AI burnout is real, and it’s shaping our collective response to AI tools. The shiny new features every platform throws at us are giving them a steadily decreasing payoff in terms of user enthusiasm and engagement.
