The Great AI Fad

Unless you have been living under a rock recently, you can’t help but have seen the near-hysterical excitement surrounding AI in the tech industry. Although the technology behind this craze has been around for several years under the banner of machine learning, the popularisation of the large language model by Open AI Chat GPT sparked this new wave of excitement and hype.

I have been interested in the tech sector for over 30 years, since I was a child and have worked full-time in it for about 17 years. I have witnessed many trends emerge and fade, from the widespread adoption of the web through the rise of the smartphone and Web 2.0 to the crypto boom and the advent of voice assistants, but apart from perhaps the dot-com bubble, I haven’t seen anything on the same scale as AI.

Over the last few years, every major and minor tech company has attempted to incorporate AI into its products and offerings, regardless of whether it makes sense or not. I don’t think there is a service you can sign up for that doesn’t feature AI as a key component. Hardware companies are incorporating AI into product names, and giants like Microsoft, Apple, and Google are integrating it into the core of their products and services.

I don’t need to tell you these things. If you are reading this, you likely understand the significant disruption that AI has caused in recent years. However, I think this one will fizzle out in time, like the trends that came before it. AI is undeniably an impressive technology. I use GitHub Copilot myself, and I don’t think the concept will disappear, but I think the hype will. I expect that AI will experience a peak similar to the dot-com bubble did in the early 2000s.

The web remains a key part of our lives 25 years later, but being online is now the norm for a business, not a unique selling point, and that is where I see AI heading. I think we peaked too early with AI; while the technology is impressive, it has deep mistrust.

Firstly, the information that AI provides can’t be trusted, as evidenced by high-profile incidents. AI hallucinates (that is, it returns incorrect information as fact) regularly, and honestly, this problem only worsens. When you have to do the work to fact-check the answer you are getting from AI, is it benefiting you? The lack of trust in AI’s output is a significant problem for AI companies, as it causes AI to be perceived as a novelty rather than a helpful tool.

Another significant issue that AI must address is legal concerns. The major AI models from big and small companies have likely been at least partly trained on data sourced illegally or whose legality is in a grey area. Generative image AI has faced massive backlash from artists and the public for stealing artists’ work. These issues surrounding legality deter professional users and corporations, which, as the highest revenue-generating users, could cause long-term harm to AI-focused companies.

In addition to the legal concerns, AI will also face regulatory challenges. Many high-profile tech, academic, and legislative leaders have expressed concerns about the potential effects of unregulated AI use on society. These concerns mainly stem from the use of generative AI. As AI improves at creating realistic video, audio, and images, it will become increasingly more complex to distinguish between real and fake content. This poses a significant risk to society, as AI-generated content could be easily used to manipulate people and seed civil unrest. These concerns and challenges suggest that AI will likely face further regulation in the future, similar to the regulation of the web.

Even with my concerns about the AI hype and the challenges it faces as it grows from a novelty to mainstream adoption and deployment, I do think, like with the web and the rise of smartphones, we are standing at the beginning of a monumental shift in how, as a species, we interact with our technology.

One risk I have not mentioned in this article is the direct financial risk. This is mainly because it is hard to quantify. Using trained models at this time holds a lower financial risk. Companies keen on establishing a user base for their models often allow use, especially for experimentation and evaluation, at a very low cost. However, developing and training a model is an expensive investment. You require high-end hardware, which comes with substantial capital costs and operating expenses. AI consumes vast quantities of energy, which, in addition to being costly, also presents environmental challenges. In the current climate crisis, this can’t and shouldn’t be overlooked.

If you are excited about AI's prospects, I encourage you to exercise caution and weigh the risks as you move forward. A significant amount of money is invested in AI, but, as mentioned in this article, I believe it is primarily driven by hype. If AI can provide tangible benefits to your business, product, and/or users, then it is worth exploring. However, as with all things, trying to find a problem that fits the solution is not sustainable in the long term.