NI (Natural Intelligence) The Creation - Next Level Artificial Intelligence

Chapter 3 The Design of the Algorithms

Months passed and Eric's place in researching the detail of his project proposal was still the library. He and Michelle frequently met there about twice or thrice a week, and occasionally accompanied by Vittoria. They still had the usual conversation but this time it was Michelle who spent more time asking and sharing her ideas with Eric. And then one day the two of them realized the attraction they felt for each other was more than just friendship or companionship. It was love.

In his second year, Eric was ready to encode his program. He spent a year researching and collecting every available piece of data, software, and information he could find, and fine-tuning his design several times. Michelle had served as his inspiration along the way, even helping him in some technicalities on cellular functions, including advanced topics in neurobiology and cybernetics. Their meetings were not just to accompany each other but to learn and share technical stuff in their respective fields.

Eric spent long nights thinking and encoding his program. He started encoding the sound processor to generally act as sound recognition system, capable of recognizing human speech and environmental sounds. Human speech, in general, had been mainly composed of the basic phonemes, the smallest distinguishable sounds. Phonemes were extensively being used by many speech recognition systems in many modern devices, such as computers, telephones, and smartphones. However, Eric combined phonemes recognition with syllabic sound recognition for double checking of spoken words. And then he programmed the sound processor to translate spoken language into stream of basic sounds, to be divided into recognizable words, and finally assembled into sentence. He added an extension quick grammar checker to double check the sentence, making sure the words were correct before storing them as discrete data. The discrete data was the basic unit of information or idea and could also comprise of two or more words. The built-in grammar checker would only function as a general feedback system to make sure the words were correct or needed to be corrected first before dumping into the temporary memory or cache RAM. He configured the sound processor so it could also access wide collection of words and symbols in an extension dictionary for fast recognition, allowing the program to readily distinguish discrete data from noise or less meaningful inputs. Once stored in the cache RAM, a discrete data would then be considered part of the program's active memory or vocabulary, and it could be translated back to spoken language in conjunction with the speech processor, which processed words to sounds. The more a person spoke with the system, the wider its vocabulary, and its efficiency was high because only the most spoken words were being stored as discrete data, thus maximizing memory usage. Eric also included an update subsystem to the sound processor to recognize unknown sounds; it would fetch the necessary attributes from the cache RAMs of other senses and create a sort of automatic learning. Furthermore, he created a subsystem inside the sound processor to help in recognizing new sounds, and he named it echolocation subsystem. The system would be used to collect the sound from two microphones (ears), and used the time difference of the sound reaching both ears to compute the location of the sound source. Echolocation's main function was to identify different sound sources to give the program capability in deciding if a person was talking to it or to the other person.

The speech processor was the next system he programmed, acting as a library of sounds for the main program where the speech engine (spoken sound of the program) was being processed. Thus, the program could speak the words it recognized using its own voice, or perfectly imitate the sounds of the surroundings, which of course considered as special talent for humans. The system worked closely with the sound processor, and could use words and grammar checking. To give more flexibility to the spoken sound so it would not sound monotonous or robotic, Eric incorporated a modulator program for the spoken phonemes or syllables. The modulator program could modulate the sound, creating variation up and down tonality as well as long and short syllabic pronunciation. The modulator had built-in intonation and the capability of learning or imitating human intonations. The speech processor compiled the sounds before speaking them out to imitate humanlike responses, even imitate the process of vocal singing.