“It takes a village…” (An old African saying)

MaMuTh is a concept to be applied to any system available to a public of participants. It is an interaction tool for a virtual entity, a “bot”, which is augmented by the incessant scripting of usage. The key factor of MaMuTh is the feedback interpretation, building on past generations of educational archives. MaMuTh improves its “speech” (which doesn’t necessarily have to be textual or vocal) by practicing conclusions derived from chatting “experience”. Instead of trail-and-error, MaMuTh will not attempt to execute responses it did not encounter on history files, at least not on the first stages. instead, it will learn to recognize component types, produce reasonable syntax combinations, and deduce the underlining logic behind it all.
The front-end can be manifested in any fashion containing the exchange element. The basic presentation is an input text-field, where a sentence is written and submitted to MaMuTh to be replied. Nevertheless, the two-way backtalk principle can trade not only a sequence of words, but also graphic symbols, user-preference qualities, tagged pictures, or any other variable values to match and assemble together. Also, it can be assimilated into grander complexes -games, web-sites, IMs, online social groups and forums… separately, or as a branched out unit.

MaMuTh agenda has two prospects (associated but independent):

Primary: Reaching local familiarity to the environments it is posted at. The ideal is not to eventually result in a comprehensible articulation. MaMuTh’s ultimate goal is to acquaint itself with indigenous vocabulary and shared tendencies to certain arrangements, instead of others. The desired effect is a meaningful (in all sense of the word) relationship between MaMuTh and its users. By adjusting to iterative verbal (or any other sort of constituents) patterns, with each session it re-calibrates to stay relevant and harmonic. As for being exact and correct – as long as it is understood, that would be “close enough”.

Secondary: Large-scale statistically stacking of records, containing authentic protocols, attained from the scattered interfaces (or rather one edge). It is an open-ended project, intended to revise not only the limited scope of immediate communication, but to isolate core truths and general rules arising from all the evidence. The progression will also be directed and dedicated to a continues examination and alteration of the methodical foundation.

The MaMuTh’s actual implementation is made up of two key sections (each may be posted without the other):

1. The MaMuTh Accumulator – an interconnected parent framework where all the gathered information is documented and stored. The MaMuTh Accumulator (“MAAC”, henceforth) has many agents. Some are different services formulated on the MaMuTh development, while others are extensions of the original engine. In any case, all their activities are dispatched to the source, wherever it may be established. It should be noted that the “MAAC” can function on its own, without occupying a net, as just a single apparatus.
2. The MaMuTh Mind – Residing on the collective server where all the data is transfered to, The MaMuTh Mind (“MAMI”, henceforth) is the part responsible to process all user-to-program dialog that had been taking place on any of the aforementioned MaMuTh agents (“MAAG”), and in turn – sent to the “MAAC”. Its job is to eventually analyze the raw conversations (lingual or not) by utilizing advanced didactic procedures. It is, in fact, a carefully woven algorithm, purposed to perpetually refine the dialect capabilities, accuracy-wise and flexibility-wise. The distinction being a separation between:
A. The persistency with the common human perception, universal grammar rationalization, and phrasing meta-models.
B. The versatility of possible cultural and demographical slang, personal expressive styles and tempers, and exclusive individuality.

Expected Issues

In order to end up with a viable platform, several obstacles need to be surpassed…
MaMuTh will have to be launched with an initial body of knowledge configurations (e.g. ciphering skills or simple cognitive mechanism). It mustn’t start from point zero, or else a nullification loop would not allow any evolution to happen. Furthermore, introduced to the empty vessel, users will only receive blank answers.
MaMuTh cannot do without content. The subject character will not be interesting otherwise, and there would be no incentive for returning visitors. Later in “life”, it can rely on previously accessed materials from pioneer contributers.
During the course of surveying, there would surely be incidences when MaMuTh calculates contradicting signals, or the occasional misconduct. Its influential juvenile features, could make that dangerous. Since MaMuTh is a communal instrument, nurturing it involves politics and a code of ethics.
The above is suppose to be possible due to similar research endeavors already tested and proved successful. Spam is not a new problem, nor AI training. MaMuTh will harness the existing achievements and use the power of the massive multiuser feedback to try to accomplish what a two year-old effortlessly does, but computers still have difficulty with. But more importantly – be a good friend, even if just simulated.