The Blurring Lines of Artificial Intelligence: What Really Counts as AI?
In today's tech-savvy world, the term "Artificial Intelligence" (AI) is ubiquitous. From startups to global enterprises, the buzz around AI promises revolutionary changes in various sectors. However, a closer look reveals a perplexing trend: many of the technologies hailed as AI today were known by different names just a decade ago. Tasks performed by Optical Character Recognition (OCR) systems, once categorised under basic automation, are now often branded as AI. This shift begs the question: what really counts as AI?
To explore this ambiguity, I reached out to several experts and academics specialising in computer science and AI. Their responses underscored the evolving and often subjective nature of what constitutes AI.
One expert remarked, "AI is whatever people perceive it to be." This perception-based definition highlights the fluid boundaries of AI, influenced by rapid technological advancements and public understanding. As new capabilities emerge, the scope of AI seems to expand, encompassing both innovative breakthroughs and refined versions of existing technologies.
Another professor offered a more technical perspective, defining AI as "an automated algorithm that has the ability to learn or improve its utility over time." This definition aligns with the traditional understanding of AI, focusing on machine learning (ML) and deep learning algorithms that enhance their performance through data exposure and iterative improvements.
Yet, this raises a critical point: why are tasks like OCR, which involve recognising text from images, now considered AI? In the past, OCR was regarded as a straightforward application of pattern recognition and rule-based systems. However, with the integration of machine learning techniques that enable OCR systems to handle a wider variety of fonts and languages with higher accuracy, the line between basic automation and AI blurs.
The "Robo-debt" scandal in Australia, which involved the use of automated systems to incorrectly identify and pursue welfare overpayments, has sparked significant controversy. Some commentators have been citing the Robo-debt fiasco as a warning to users of AI. However, Robo-debt relied on basic automation and flawed algorithms, lacking the sophisticated learning and adaptive features characteristic of true AI. This misuse not only undermines public trust in automation but also highlights the need for precise terminology and ethical standards in deploying such technologies. Conflating Robo-debt with AI underscores a broader concern about the misuse and mislabeling of emerging technologies.
The marketing advantage of branding a product as "AI-powered" cannot be overlooked. Businesses leverage the AI label to attract investment, spark consumer interest, and gain a competitive edge. This trend reflects a broader cultural and economic phenomenon where the AI label is often used to signify cutting-edge innovation, regardless of the underlying technology's complexity.
The consequences of this rebranding are twofold. On one hand, it democratises AI, making advanced technologies more accessible and understandable to the general public. On the other hand, it risks diluting the meaning of AI, potentially leading to unrealistic expectations and confusion about what AI can and cannot do.
The definition of AI is increasingly becoming a matter of perception, influenced by both technological progress and societal trends. Whether viewed as a marketing tool or a technical term, AI's identity is continually reshaped by how we choose to understand and apply it. As the boundaries of AI evolve, it would be wise to maintain clarity about its capabilities and limitations, ensuring that the label serves as a meaningful descriptor rather than a mere buzzword.