In recent months, pages across the internet have been adding a symbol of two or more four pointed stars. These symbols represent AI features, which are becoming more and more prominent but at the cost of software usability and user experience.
Since AI’s introduction to the general public as a rudimentary tool that could create blurry images that vaguely represented text prompts, the technology has grown exponentially in its abilities. Programs such as ChatGPT and Midjourney have introduced advanced tools that rely on the software to generate entire stories and images within moments.
This dramatic advancement in technology is due to the progress in developing large language models. These LLMs are trained off of millions of pieces of data that come in a variety forms, a process that requires a substantial amount of processing power in order to function. The technical requirements behind such feats of software development lead to significant costs associated with both training a model and running it.
Rising costs have necessitated enormous amounts of investment opportunities, which has been readily supplied by eager venture capitalists that see AI as the “next big thing.” Such investments have propelled companies like Nvidia to new heights as their stock value has more than doubled since October 2023. Nvidia, a computer hardware manufacturer that creates the processing equipment used to run LLMs, has become one of the three most valuable companies in the world.
The meteoric growth, however, has begun to slow down in recent months. Nvidia’s stock price is no longer growing at the rate it was only a year earlier, so have the stocks of other AI-related companies. Even Goldman Sachs, a prominent investment bank that was initially enthusiastic on the implementation of AI into a variety of services, is now expressing doubt as to whether LLMs are as effective and worth the costs as initially perceived.
In an interview with The New York Times, Jim Corvello, the head of stock research at Goldman Sachs, likened AI to just another technology bubble, like cryptocurrencies. Speculative bubbles have been a hallmark of the Silicon Valley culture since its inception as new innovations occasionally gain popularity and are espoused as the next great innovation, only for the excitement to collapse in a matter of years.
The Dot Com bubble of the late nineties and the more recent cryptocurrency bubble have both led to fallouts throughout the markets and technology as a whole. Such technologies were overpromised and overvalued, and they ultimately underdelivered, leading to investor panic and most related companies collapsing.
The cycle of companies overpromising on the capabilities of technology and underdelivering is detrimental to the progress of technology as a whole, and the AI bubble is no different from its predecessors. AI technology is built on a foundation of stolen work from artists and content creators from across the internet, creating both ethical and legal concerns as to the practices involved. The technology is also energy intensive, requiring terawatt hours of energy to support the currently existing AI infrastructure.
Alongside the technical struggles that accompany AI technology are the lack of practical benefits in most implementations. AI has been added to a variety of websites and software in counterproductive ways that ultimately worsen the user experience, such as the replacement of the Instagram search bar with Meta AI. Forced introductions do not benefit the consumer, but instead, they worsen the website’s usability through adding unnecessary hurdles and features.
Although AI can be a useful tool for a variety of applications, such as in prototyping and workflow optimization, its overvaluation and unnecessary inclusions in applications are inevitably going to lead to the AI bubble bursting. The repetitive cycle of technologies being introduced as “revolutionary” is harmful to the consumer and will always yield the same result, a result that the AI market is starting to experience for itself.