Introduction

For the past few years, I have been witnessing, firsthand the fallacy that has been sold as AI. I have been trying to help demystify AI for small businesses, so they understand how to use it or exist without it. I have recently read the article from MIT regarding the failure of AI and the promises that came with it. So I thought I would write this blog to share some of my findings. Would love to hear what everyone else thinks about it and if they have a different, and hopefully nonbiased view.

 In the last decade, artificial intelligence has soared to the peak of technological advancement, promising breakthroughs that would transform every facet of human activity. From the excitement surrounding self-driving cars to the expectation of super smart digital assistants, AI has been heralded as a revolution without precedent. Yet, as the dust of false claims begins to settle, a sobering reality emerges. Much of the AI hype is failing to translate into the tangible, world-changing progress that was once promised, both in business and our daily lives.

The Origins of AI Hype

To understand the current disillusionment, one must first revisit the origins of AI’s sudden rise in public and industry perception. The beginning of the AI hype can be traced to the early 2010s, when advances in computational power, access to massive datasets, and new training techniques for neural networks led to astonishing leaps in machine learning. Achievements in machines beating humans in games, and the rise of voice assistants like Siri and Alexa, gave the impression that general artificial intelligence was just around the corner.

The media cycle and large corporations, ever hungry for the next technological marvel and increases in sales and profits, amplified these achievements, sometimes blurring the line between what AI could do and what it might soon accomplish. Venture capital poured into AI startups, and corporations raced to brand themselves as leaders in AI adoption, fearing irrelevance in the face of disruption.  The hype reached such ridiculously high levels, that world leaders as well as tech giant leaders gathered in Davos to understand how to regulate and control AI not to completely destroy world economy.

All that fell apart very quickly!!

Promises Made, Promises never delivered

The promises were extraordinary: self-driving taxis and trucks would render human drivers obsolete, AI-powered doctors would diagnose disease with superhuman accuracy, and intelligent agents would manage everything from our calendars to our economies. Governments, corporations, and citizens alike were swept up by the promise of a future made frictionless by artificial minds.

But beneath the hype, key limitations persisted. Self-driving technology, once predicted to be widely available by the late 2010s, remains mired in regulatory, ethical, and technical challenges. Very similar to adoption of EVs, which was supposed to eliminate all ICE engines by 2019, the lack of infrastructure hampered the growth and expansion of AI. While AI can diagnose certain diseases from images with high accuracy in controlled conditions, real-world applications are full of uncertainty, data biases, and lack of interpretability and explainability. The digital assistants that were supposed to anticipate our every need still stumble on context, nuance, and ambiguity. The gap between the promise and reality has become increasingly noticeable.

Why the Hype Is Failing

The unraveling of AI hype can be attributed to several factors:

Overpromising, Under-Delivering

In the pursuit of funding and attention, many stakeholders overstated AI’s capabilities and timelines. Ambitious claims created unrealistic expectations, against which the incremental, sometimes invisible, progress of real AI research appeared underwhelming. Fear of failure, added to the problem, where failures were covered up and not reported!!

Technical Constraints

AI’s most impressive features are, in fact, narrow applications, so-called “narrow AI”, that excel in specialized, well-defined domains. These systems falter when confronted with the unpredictability and richness of the real world, revealing the gap between narrow AI and the mythic “general AI.” Problems such as explainability, data privacy, and algorithmic bias continue to elude easy solutions.

Data and Scale Limitations

The performance of AI often hinges on the availability and quality of data, as well as the domain knowledge on the business side . In many fields, data is messy, incomplete, or protected by privacy laws, limiting the applicability of AI solutions. Furthermore, the business owners lack of knowledge empeeded th ability to explain the “ask” to data scientists who were supposed to crate the train the models, rendering the results unusable. Moreover, scaling up AI systems from POC and Pilot stage to fully function business systems exposes unforeseen challenges, such as high costs of compute and model maintenance by expensive platforms and data scientists, as well as adversarial attacks to ethical dilemmas.

Economic and Social Realities

Many of AI’s most-advertised applications confront not only technological but also economic and societal barriers. The job displacement predicted by AI has been slower and less sweeping than anticipated, while legal liabilities, social trust, and cultural attitudes complicate adoption.

Media and Marketing Falsification

The narrative of inevitable AI dominance is perpetuated not just by technologists, but by marketers and journalists who benefit from sensational stories. This feedback loop inflates expectations, making even genuine breakthroughs appear as disappointments when they fall short of exaggerated forecasts.

General AI and Consciousness

The notion of artificial general intelligence, the idea that a machine could match or surpass human intelligence in all domains, remains firmly in the realm of speculation. While large language models and generative AI have made astonishing progress in mimicking aspects of human language and creativity, they lack genuine understanding, common sense reasoning, and the capacity for independent thought.

The fact that humans need to “train” and “Maintain” models proves the non-intelligence of the so-called AI. Most of the AI solutions have not passed POC or pilot stage, due to a few facts:

• High cost of computing related to running the models.
• High cost of Data scientists required to constantly maintain, improve and explain the results of the models
• Lack of trust in results, requires comparison to and validation by human (Excel) based results, which defeats the promises AI made that it will reduce our workload!!

The Human Cost of Disillusionment

The failure of AI hype carries real consequences. For workers, the constant drumbeat of automation anxiety can create needless stress and uncertainty, and cause eventual disconnect. For organizations, chasing the latest AI fad can waste resources and distract from more meaningful innovation. It also creates fear in the incumbent staff who also think AI will replace them and creates resistance to adoption.

The focus has been on adoption of AI, so companies don’t fall behind the industry and become outdated. It has also been the goal of a lot of IT leadership to make sure they have AI implementation and success  on their CVs, to secure the next big job. This has been detrimental to the business units who continue to crave the advanced analytics results they were promised to grow sustainability and stay competitive.

Toward a More Responsible AI Narrative

AI is not failing in the sense of being useless; rather, it is the narrative around AI, the inflated expectations and premature declarations of victory, that is failing. To move forward, it is crucial to recalibrate our relationship with this technology. Policymakers, technologists, journalists, and the public must strive for a more realistic understanding of what AI can and cannot do. We also need to focus on timelines. AI may, in the far future, achieve what this round of hypes have promised, but not in the short timelines that fear mongers have published, in order to make $$$. Furthermore, I don’t think AI can be achieved with the current binary chipset we have. Yes, we are able to fake the illusion of “intelligence”, but we need different type of compute power to allow for Machine learning and self maintaining models.

We should still celebrate the real progress, even if incremental, and acknowledging limitations without resorting to doom or euphoria. It means encouraging transparency and accountability in research, resisting the temptation to oversell, and investing in education that demystifies AI.

Conclusion: Lessons from the Hype Cycle

The story of AI hype and its failures is not unique; it echoes earlier cycles of technological enthusiasm and disappointment. Outlook has been dying since 2010, Excel and a list of other technologies have been doomed to die for decades, yet they thrive and grow into our daily business and personal lives. Despite all of this each cycle offers an opportunity to learn, to recognize the dangers of overpromise, and to foster a more balanced, sustainable approach to innovation, with more realistic timelines.

The so called “Artificial Intelligence” remains a powerful tool for automation and mining hidden insights from massive amounts of data we have accrued in the past decade, capable of reshaping industries and societies in profound ways. But its journey is one of evolution, not revolution, of gradual mastery over complex problems, rather than sudden leaps into science fiction. By tempering hype with humility, we can ensure that AI’s brightest days still lie ahead, built not on fantasy, but on the enduring promise of human ingenuity guided by reason and responsibility.