The Concept Of Attention Is All You Need
A human mind, teaching itself through encoded media, works the same as deep learning encoding digital input for a transformer. However, this is only when learning happens from restructuring the brain to deal with distortions in a natural language delivery system (writing) to overcome a disability (dyslexia) by making the individual’s attention, alone, the decoder.
Those severely dyslexic, classified as overcoming the condition, have higher than normal reading comprehension. The coding/decoding from checking and rechecking the difference between what was intended to be comprehended and the meaning non-dyslectics only find logical in the syntax and context of the language, adds this depth to what the dyslectic learns.
The most capable training/testing procedures for neural networks use a process with a real affinity to the same approach to reasoning. Analysis of meaning from what natural language finds the attention of the reader finds recognizable as indicators of what makes sense, however, does not support the same level of intelligence.
The realization that the relationship between attention to transforming another’s data into an individual’s awareness of the external world, and the attention that the dyslectic falls back on in experiencing the data as several logical states of being true or false, means examining only attention to the context of the input’s meaning will always lead to a fallacy in learning.
That state of awareness will not be what makes Artificial General Intelligence a reality. da Vinci, Napoleon, Churchill, Einstein, Jung, Edison, and the like, all turned dyslexia into uses that made their depth of attention into data that is now found to be giving transformers accurate biases, not creative responses.
Give anyone an exam to see if any symptoms of dyslexia are there, and something shows up. It’s natural to rethink input. The more that the human mind must make sense of the accumulation of input from the enormous digital resource gathering on the Internet, the more there is to transform. What appears a natural working state of an active mind to a functionally dyslexic learner observing GPT3 output is what’s the goal of AGI, not attention recognition responding to an algorithm that some conformist has made for baiting a performance.