By: Margarita Garcia - Managing Director, Naoitech
The current narrative that implies that we have achieved AGI or that LLMs or AI Agents are somehow plotting to destroy humanity is not based on technical or scientific explanations but rather on a cultist belief. When proponents of these theories are pressed for explanations, their conjectures always lead to fantastical notions that require an actor in the physical world executing these master plans; their premises always require human intervention or are problems that can be solved with proper guardrails and adequate cybersecurity.
There are very specific areas that truly should concern us about Generative AI, such as data centers depleting our natural resources in the race for the best LLM, the mental crisis that is leading many people to AI psychosis, the dependence of the economy on the AI bubble, and irresponsible companies using Generative and Agentic AI in high-risk industries. However, one of the most dangerous new threats is the cognitive decline that the overuse and addiction to chatbots are creating.
The paper "Large-Language Models as a Cognitive Virus" studies that very issue. The Researchers explored the damaging consequences of overusing chatbots in our daily lives and how that could lead to a decrease in cognitive competence.
"Large language models are rapidly becoming embedded in everyday cognitive and cultural practices, yet we still lack simple theoretical frameworks for understanding how their adoption may reshape collective patterns of cognition. Here we introduce the idea of LLMs as cognitive viruses: culturally transmitted technologies whose spread is promoted by their usefulness but can also increase dependence through cognitive offloading. A dynamical model shows that social facilitation can generate runaway transitions from predominantly autonomous cognition to widespread technological dependence. These results suggest that abrupt collective changes can emerge even when individual adoption is gradual. By connecting epidemic dynamics, cultural transmission, and cognitive offloading, our framework identifies general mechanisms that may govern human–AI coevolution and provides a basis for “cognitive immunization”.
Can a robot become a good Chef? The answer is categorically NO. The explanation is quite simple: robots don't have taste buds, so they are not able to understand the difference between sweet and salty or to comprehend how to achieve the elusive umami flavour. Let's assume that the robot confuses the salt with sugar or loses its ability to perceive quantities; how will it be able to discern that the recipe is wrong? It can't; inanimate objects, LLMs or cloud software can not perceive or understand the physical world; that is why humans will always be superior to any machine, no matter the Jules Verne-inspired rhetoric that points otherwise.
Offloading all our knowledge, creativity and innovation to an LLM or cloud software which circles the same data is not going to bring any new ideas, concepts or creative abilities into the world; in fact is going to bring an extremely boring and undiversified future. However, knowing how to use technology to develop new concepts and ideas while maintaining our ability to create and innovate can bring about real change in our society.
The oversimplification of human complexity is harmful, the same way that a decade ago the social aspects of humanity were transformed into digitally isolated, harmful interactions with strangers who commented on areas that they have no prior knowledge or understanding. If we take the lessons from the social media debacle, we can mitigate some of the risks that the Gen AI era presents.
Humanity must preserve, train and advance the gifts that make us unique while understanding that technology has a significant role in helping society when used responsibly, safely and in its proper context. We are the masters of technology, not the other way around.
Sources:
The AI Resistance is Forming- Cal Newport podcast
Large-Language Models as a Cognitive Virus https://arxiv.org/pdf/2609.03344
https://www.sciencedirect.com/science/article/abs/pii/S1364661316300985