The Logic of Language

The field of Artificial Intelligence (AI) and Natural Language Processing (NLP) — in a broad sense — has fundamental problems since its start because it is lacking a foundation in nature:

​​

• AI is lacking a unifying, fundamental (=natural) and deterministic (=implementable) definition of intelligence;

• Regarding NLP, scientists have no clue how the human brain organizes its knowledge. As a consequence, Large Language Models are unable to organize changes in their knowledge structure independently.

​
Not a single variant of the Evolutionary hypothesis of Common Descent provides a profound understanding of natural intelligence and natural language. So, scientists have no better option than to mimic behavior in the field of AI / NLP, using behavioral/cognitive science. However, mimicking a hen’s (chicken’s) behavior will not produce a single egg.

​

Expressed as another comparison, AI / NLP is a flight simulator rather than an airplane: A flight simulator moves pixels on the screen — and it moves the cones of the speakers —  but it will not move an inch.

 

Put in even different words:

• If one creates a logical (=intelligent) system to process a natural system — like, e.g., the human language — one will end up with Artificial Intelligence, which is just a (mis)interpretation of that natural system;

• If one considers a natural system — like, e.g., the human language — to be a logical (=intelligent) system, one will end up with an artificial implementation of natural intelligence.

Artificial Intelligence (AI) may deliver useful engineered techniques. However, engineering is limited to delivering specific solutions to specific problems, while science delivers generic solutions based on the way nature works.

​

AI is limited to human-written, human-selected, and human-programmed intelligence. As such, AI will never become naturally intelligent. In general, AI is limited to performing repetitive tasks. Artificial neural networks — in particular — are limited to pattern recognition and pattern generation.

My father taught me: “Don't become a monkey who is trained to perform a trick”. Artificial neural networks do not learn anything. They are just trained to perform a trick. Neurons are not even essential to intelligence, in the same way as feathers and flapping wings are not essential to aviation.