The underestimated risk of AI

In the ongoing discussion about artificial intelligence (AI), it seems that one crucial risk is often overlooked or, at the very least, not fully appreciated. I am not referring to Artificial General Intelligence (AGI), self-awareness, or Singularity. While these topics are important and worth discussing, there is another risk, more tangible and more tangible and closer to us in time. I am talking about something as simple as the effectiveness of AI algorithms.

In my argument, the key is not algorithms' "intelligence" level but their "effectiveness" – the ability to achieve the intended outcome. The fact that algorithms perform their assigned tasks better, faster, more accurately, and more cost-effectively than humans brings obvious benefits but also entails risks.

Humans, by nature, lean towards comfort, laziness, and optimization. When algorithms operate in line with these natural human inclinations, they become increasingly used and treated as something obvious (a commodity). Personally, I cannot imagine travelling to an unfamiliar destination without using GPS. What does this imply? It means that we are increasingly relying on algorithms, losing certain skills and independence in the process. I used to navigate terrain subconsciously, remember roads, but now I rely on GPS. This tool has replaced certain skills, weakening my ability to navigate independently.

A similar situation could be observed after the invention of fire at the dawn of humanity, which brought comfort but also led to the loss of certain skills, such as vigilant sleeping or preparing raw food. Now, in a hybrid reality where the "virtual" element is rapidly growing, relinquishing power to algorithms may lead to a loss of control, both to their creators and the algorithms themselves.

Giving decision-making (and in practice, this means relying on their effects) to algorithms in the first place may actually mean giving control to their human creators. Over time, as algorithms "liberate" themselves from human control (human creators), the power may shift to the algorithms themselves. In this perspective, a realistic scenario (relatively near) is the "domestication of humans" by algorithms (like in “domestication of cattle”), potentially leading to complete predictability and manageability of their (humans) behavior. This, in turn, raises questions about free will and consciousness. Do these values truly exist, or are they mere illusions?

My point is clear:

The performance of algorithms and their widespread use carries significant risks to humanity regardless of its level of intelligence.

The discussion on this topic is becoming increasingly urgent as awareness of the consequences of over-reliance on artificial intelligence is needed. We must find a balance between exploiting the technological potential (using the efficiency of algorithms) and maintaining our unique capabilities and independence. Otherwise, we risk losing the values that define us as people.