Intelligence: Between Neurons and Algorithms

Intelligence is a fascinating and complex theme that has long captivated philosophers, psychologists, and biologists. In recent decades, engineers and computer scientists have also attempted to replicate it in the electronic field. However, defining what or who is "intelligent" is far from being resolved. As highlighted in the introduction, intelligence is a concept characterized by the lack of an objective reference point. This means that there is no precise, measurable, and definable criterion, but rather a series of characteristics or canons that suggest the tendency to associate and process events or information towards efficient problem-solving.
In this sense, intelligence can be abstractly identified as the ability of a system to compute information in the most complete possible way, namely by calculating as many variables as possible in a given environment in an efficient and creative manner. This allows for the resolution of one or more specific questions or, mathematically speaking, the identification of unknown data starting from known data, adapting this method to various unknowns and learning from past experiences. Therefore, the very idea of intelligence cannot be reduced to a simple direct comparison of available means or values, but requires a capacity that goes beyond linear processing towards what can be defined as "heuristic", a term derived from Greek meaning "to find" a solution to a problem.
To better understand what is meant by heuristic processing, it is useful to make an example from everyday life: when we face an unexpected situation, our mind does not proceed automatically, but activates strategies for prediction and supposition aimed at solving the problem through alternative associations and experimentation. This processing mode represents an important characteristic of the very concept of intelligence.
From a biological standpoint, this is a property that has developed over the course of evolutionary development of organisms in conjunction with the physical ability to adapt to the environment and interact with it in the most efficient way. Current studies define the level of efficiency of this interaction with complexity and the number of synaptic associations that allow for learning new information beyond genetically innate information. In more developed organisms, this has led to verbal communication, emotional empathy, and complex emotions.
From an informatics perspective, the goal of creating systems capable of processing and solving problems has been analyzed since the early days of computing. Without going into too much detail, it is only in recent years that the calculation capacity of computers has been able to be applied in human-like tasks, but in what way? It should be noted immediately that current algorithms are primarily based on statistical learning and comparison, i.e., they are deficient in self-determination, which means they are unable to go beyond their intended purpose or orient themselves towards human models, not because of complexity but because of creativity.
In other words, current artificial intelligence systems are essentially at "zero heuristic advancement" compared to the previous criteria described. This does not mean that these tools are useless or unperformant, but rather they can express their best performance in specific contexts, namely those that require large-scale processing of big data. Since it is impossible for a human system to achieve such mnemonic elaborative speed, artificial intelligence is a highly valuable aid that excels in certain fields and is even superior to human comparison in certain areas. However, it remains subordinate to biological flexibility.
This observation has important implications for the future development of artificial intelligence, which will need to overcome these limitations to approach the complexity of neural and synaptic functioning in biological brains. There are those who believe that with progressive increases in calculation power, such flexibility will be emulable at a level where no differences will be noticed. If we want to achieve this goal, algorithms will need to at least emulate the ability to process heuristic information and exponentially increase the power of artificial neural networks using evolutionary algorithms based on principles of artificial intelligence.
Moreover, there are schools of thought that associate consciousness and intelligence on an existential plane: according to this view, consciousness - namely self-recognition of one's existence - is a condition (if not sufficient) necessary for intelligence to exist. Against this last affirmation, artificial intelligence does not set a purpose or goal; therefore we are far from achieving "artificial self-awareness", although it is one of the most recurring themes in science fiction or popular culture. However, it remains an premature hypothesis given the complexity and specificity of human neural structure.
It could be ventured that a combination between advanced heuristic behavior and self-determination - namely the ability to set goals - along with curiosity and desire to understand what surrounds us - might be what we should replicate in machines if we want them to develop a type of intelligence like human intelligence. This would require creating an artificial neuron with a 1-to-1 relationship with its biological counterpart, an objective that we are currently far from achieving. However, science and research continue to advance exponentially compared to past progress; therefore this target may not be so utopian after all.