What ChatGPT’s downturn tells us about AI adoption

ChatGPT witnessed its first downturn in monthly active users since its inception.

As the hype surrounding generative AI trends downwards from its meteoric highs, and ChatGPT sees its first downturn in monthly active users since its inception, the moment warrants some sombre reflection on the state of AI adoption today. In the recent past, extreme hype had led to a complete decoupling of expectations from reality, with many exaggerated fears of mass-layoffs and population-level unemployment being freely thrown around.

However, this tsunami that we were warned about, as being imminent, is yet to present itself. And, for the time being, things are poised to continue this way. This is because, despite the double-exponential growth that AI has seen (as stated by Elon Musk), there are several inhibitory factors that continue to forestall the widespread adoption of AI.

Now, before any discussion on the nature of these restraining factors which are hampering widespread AI adoption, some stock-taking should be done as to where we stand today. This, surprisingly, is a particularly difficult task for many reasons.

AI: Everywhere And Nowhere

First, AI is a catch-all term that encompasses everything from the recommendation algorithms of YouTube to those predicting your next word on a Google search, to those reading and analysing research papers for cancer treatment. Therefore, when we talk about AI, it is not obvious that two parties are referring to the same thing.

Second, due to the existence of this definitional ambiguity, many firms have thrown around the term indiscriminately. This is part of their strategy to look future-ready and appear more palatable to investors.

We can see this phenomenon playing out by looking at the findings of MMC Ventures, which shows that over 40 percent of nearly 3,000 “AI firms” in Europe don’t use any machine learning at all. Their findings also show that AI-focused startups receive 15 percent more funding on average, showing that a perverse incentive exists for firms to market themselves as AI adopters & innovators.

The next dimension to explore is the pattern of AI utilisation. Of course, just having AI functionality available within the organisation is not enough. It needs to be actively leveraged, and its broad functionality utilised.

However, like the above factors, here too the picture is rather lacklustre. Accenture, in its 2021 survey, found that only 12 percent of the world’s richest 2,000 firms are using AI at a mature level, i.e., using AI in a way that gives them a competitive edge.

63 percent of firms surveyed were only AI experimenters, barely scratching the surface of what was possible. While this survey is two years old, the picture it portrays is unlikely to have radically changed so far, given the lack of nimbleness at large organisations and the pandemic induced struggles in the interim.

Overall, we can see that the term itself isn’t clear, how many are actually using it isn’t well quantified, and the way in which it has been deployed continues to be largely unsophisticated. Now, with this context in mind, finally, we may begin to postulate why the adoption rate might be lagging in the manner in which it has.

Not A Must-Have Tool Yet

The first answer to the above, prosaic as it may sound, might be simply because of an unawareness among people about AI’s utility – both on an individual and organisational level. To illustrate, UK’s Office for National Statistics found that while most adults know about AI, they don’t find it relevant to their work. It shows that use cases simply have not been explored sufficiently, and the nascency of the technology in the public imagination may be a reason for this.

A second reason for lagging adoption could be fear and wariness. And many factors could be contributing to this. To begin with, generative AI cannot show you how it reached a specific output. It doesn’t tell you its assumptions or give you any disclaimers.

Its reasoning cannot be traced step-by-step, due to which even its founders are often unaware as to how it works. It cannot always cite sources correctly, and it often hallucinates by making up information that is plainly false.

Neither is the quality of engagement consistent, beyond a certain number of prompts. None of these weaknesses do its image any favours.

Even the output of generative AI is not consistent like regular code. As the system evolves and learns from its interactions, its performance on different tasks varies. It is possible for AI systems to actually become worse over time, as was recently witnessed in the case of ChatGPT. This is not particularly endearing for AI, from a business perspective, as output inconsistency is something that most executives will not be able or willing to accept.

The fourth reason for wariness is data. Large AI models are trained by a few firms, with data of their choosing. This data can have inherent biases, and it may contain copyrighted data that it accessed illegally. Even engagement with AI systems introduces the possibility of data breaches and leaks.

Businesses, before accessing these large models via APIs, need to be certain that these biases won’t manifest themselves in the output delivered by generative AI, and that copyright infringement will not lead to myriad legal issues. This is a tall order for any firm, given the nascency of the technology, and the unexplored nature of AI’s benefits.

AI Wariness Is A Good Thing

KPMG’s 2023 survey which finds most adults in 17 leading AI nations being wary of the technology is a direct consequence of the above factors.

Thus, we can see that various factors have subdued the rate of AI adoption, beyond the ever-present cost and skill barriers. Due to this, the dramatic future that many warned us of, of an AI induced upheaval of society and economy, appears to be a fair distance away. Interestingly, this sluggish pace of adoption might actually turn out to be a good thing.

This is because it gives the requisite breathing room for business, legislators, and civil society to push for a well-rounded and well-regulated AI ecosystem. And that could just be the silver lining that the world needs right now.

Srimant Mishra is a computer science engineer from VIT university, Vellore, with a deep interest in the field of Artificial Intelligence. He is currently pursuing a law degree at Utkal University, Bhubaneshwar. Views are personal, and do not represent the stand of this publication.