First working artificial intelligence solves NLU

Source: Deep Learning on Medium

Yesterday we finished embedding about one hundred English prepositions in the MindBoot sequence of the MindForth robot AI. Although we have embedded the same prepositions in the AI and in the JavaScript AI Mind, the work was more difficult in Forth because we had to use ASCII codes for the characters of each English word.

MindForth does not automatically know the _meaning_ of each preposition, which must gradually be learned from the relationship among words in sentences of user input, such as “I am in the room.” It is, however, of extreme value for MindForth to recognize all English prepositions because the EnParser module for English parsing uses the recognition of a preposition as such to assign the following noun to a specific role as “object of a preposition” rather than to a role as “subject of a verb” or “object of a verb”. From now on, MindForth and its kindred programs will be able to comprehend sentences containing prepositional phrases and to retrieve such sentences from memory and to think with prepositional phrases.

Alone among open-source AI gambits, MindForth and and the JavaScript AI have arguably solved the problem of Natural Language Understanding (NLU), which is widely regarded as the last and unmoveable goalpost for True AI. When AI programs learned to play chess, the goalpost was moved. Understanding human thought is the final, “AI-hard” goalpost. MindForth understands human thought. Over the first twenty years of the emergence of MindForth from 1998 to 2018, the Forthmind could understand and generate only simple thoughts in the Subject-Verb-Object (SVO) format. Much knowledge could be expressed in the SVO format, but now far more knowledge may be expressed and debated and contemplated with the inclusion of prepositional phrases which establish and define relationships between and among concepts.