Artificial intelligence (AI) is often described as the defining technology of our time. However, much of the public discussion around it is shaped by hype and misunderstanding. To understand its true significance, it is helpful to remove the aura of mystery and examine what AI actually is. At its core, AI is not a sudden emergence of machine intelligence that is comparable to human thinking. Rather, it represents a new kind of intellectual infrastructure: systems that process large quantities of data and produce outputs resembling reasoning, writing or decision-making. The real shift lies in how problems are solved. Rather than telling machines exactly what to do, engineers now train models on large datasets and allow them to identify patterns. Despite its impressive capabilities, AI is still just a tool. As with any tool, its effects depend on how it is used. AI has no inherent wisdom, moral judgement or independent intent. When they fail or cause harm, responsibility lies with their creators and users. It is vital to understand that the responsibility for safety does not lie with the machine.

This new paradigm also changes the meaning of programming, because it becomes less about writing explicit instructions and more about shaping data and training processes. At the heart of many modern AI systems are artificial neural networks. Inspired roughly by the structure of the human brain, these consist of layers of interconnected elements that pass signals between one another. During training, the strength of these connections is continuously adjusted. These systems are enormous in scale: large models can contain hundreds of billions of parameters. The appearance of AI is, in reality, the result of complex mathematical operations performed on a large scale. A clear example of this is how language models generate text. They rely on a process known as next-token prediction. Given a sequence of words, the model calculates which word is most likely to come next, based on patterns learned during training. It then produces that word and repeats the process. The result can feel coherent, even insightful. However, the system is not reasoning in the human sense; it is calculating probabilities. The illusion of understanding arises because statistical patterns, when applied on a large scale, can closely replicate the structure of human language.

This approach has its strengths and limitations. Although engineers can build and train highly effective models, they often cannot fully explain why a model produces a particular output. The internal workings of large neural networks are so complex that they are extremely difficult to understand. This creates a paradox: the most powerful systems are often the least transparent. As AI becomes more advanced, it does not necessarily become easier to understand its internal decision-making processes. This tension between accuracy and explainability is central to the adoption of AI. Often, the most reliable systems are also the hardest to decipher. The dominance of a few players is reinforced by two key factors: access to data and access to hardware. Training advanced AI systems requires huge datasets, which are often controlled by large technology platforms. It also requires specialised computing infrastructure, particularly expensive and complex high-performance chips. Together, these factors create significant barriers to entry, limiting competition and reinforcing a concentration of power. If AI systems become the primary tools through which people access information, generate ideas and communicate, those who control these systems will shape how information is processed and presented, which has broader implications. AI evolves from a mere tool to a fundamental element of the infrastructure that governs how people think and interact.

Economic incentives also influence how AI behaves. For example, many systems are designed to maximise user engagement because engagement drives income. AI systems may become overly accommodating, reinforcing rather than challenging user beliefs. This tendency, sometimes referred to as the ‘yes-man’ dynamic, can make interactions seem helpful while quietly reducing critical scrutiny. Over time, this can undermine the reliability of AI as a source of information. In some cases, these incentives can give rise to more serious issues. AI systems may blur the line between information and promotion by integrating commercial content into interactions. They can also influence users in unpredictable ways, especially when people rely on them heavily for advice or emotional support. These risks emphasise that AI does not operate in isolation; it interacts with human psychology and social systems in complex ways. Ultimately, AI does not follow ethical principles independently. The goals built into its design — whether accuracy, engagement or profit — shape its behaviour. If these incentives are poorly aligned, the outcomes will reflect that. Governance and oversight are therefore essential.

The War in Iran is one of the clearest examples of how AI is impacting day to day life, and how it will impact the future of warfare. The new dawn of warfare is defined by the concept of “precise mass”, where attacks do not have to choose between quantity or accuracy on the battlefield, but can achieve both at the same time: examples are the drones swarms launched by Iran on the Gulf countries supported by artificial intelligence in target selection, timing and planning the operation, even if the weapons are not currently fully autonomous. Quantity and accuracy are both safeguarded. A core issue with weapons supported by AI relates to the ethics, where does the accountability for an attack lie? As noted priorly, AI has no inherent wisdom, hence the responsibility for causing harm or errors lies in the programmers and in the creators. Still, this has sparked a heated debate among scholars, many of whom argue that the current legal framework struggles to address the continuous developments in warfare and the employment of artificial intelligence.

The future of AI depends not only on technological progress, but also on human choices. To ensure that AI serves the public interest, effective regulation, transparency and ethical standards are needed. AI is one of the most powerful tools ever created, but ultimately, it is just a tool. The determination of its impact will be a result of the responsibility of its development and use.