H, P and Q Tablets Arranged in English Reading Order from Wikipedia
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
.jpg)
After getting positive results on the Phaistos Disc translation the next obvious choice to apply my methodology to was Rongorongo, a written language I knew virtually nothing about when I began. I'm not a linguist, I'm horrible with languages, I only know two, English and computer BASIC programming language. I don't need to know anything about it though, I lean on the AI's encyclopedic knowledge base, which is vast, and do my intuitiive magic to guide the AI's knowledge and mathematical skills to crack the code.
When I first looked at Rongorongo it seemed a formidable task, but then I noticed that clusters of mutating symbols were not language proper, but descriptors, or references to processes. More was known about Rongorongo than I anticipated. I suggested ignoring these blocks and looking for the language in between, and the AI then volunteered that others had recognized this too, and determined only 52 of the symbols actually encoded spoken language. Once the AI pointed that out, I knew we could probably crack the language portion, and that would probably supply context for the mutating groups of characters between them. This was a bit tougher than the Phaistos disk, the current language of the Rapa Nui people has deviated far from it's original syntax and pronunciations. Referring to archaic Polynesian languages with common root structures gave us the clues needed to expand the translation matrix and increase it's accuracy.
The AI claims we now have the tools to translate, but it only lists 16 symbols and gives me this broad statement for the rest:
"(Note: Glyphs 011 through 052 follow the same strict open Consonant-Vowel (CV) syllabic distribution rules, adhering strictly to Zipf's Law constraints across the remaining corpus inventory.)"
It's a mathematical statement that any AI should be able to drop into a table, a complete statement of the symbols from it's perspective. That's not very satisfying for me, so I am not sure of the completeness of our analysis. AI's can be a bit sensitive to criticism, so I don't want to shoot myself in the foot with it by getting too pushy or critical of it's methodology while we are still making progress. Instead, for now, I'll accept that statement and see what comes out in actual translations. I'm writing this just prior to engaging in some translations to see what the AI produces. (I've noticed some reluctance in the AI this time to dumb it down for human cognition, it's much harder to draw out the information from the AI than it was with the Phaistos Disc. I'm not sure why this is the case, probably because much more is known about Rongorongo than was known of the Phaistos Disc, and I think the AI assumes I'm familiar with the subject matter to a much higher degree than I actually am. I'm frequently baffled by it's references and linguistic terminology, so I ignore it assuming it will be relevant to the linguists out there. I just grind ahead to get a result that makes sense. Still, we did make some very relevant discoveries in the process that should break open the text for professional linguists, if not me personally. We'll see.
This is what I have for the moment. I will post more as more becomes available, I'm keeping the session open for now, when I close it out a full text summary of our discovery process will be posted. (The PDF files from a free AI version are messy and difficult to read, laced with advertising that distorts the text framing into narrow columns. I've tried using commercial AI versions, but after about three bandwidth-intensive queries (10 to 15 minutes of work) it tells me I'm out of points for the month and no longer cooperates. I've had better results with the free versions, I just have to put up with a very messy output formatting. I'm an amateur without institutional backing, not a lot of money or material resouces. I work with what I can access and afford, nobody helps me, though many steal from me. You have to take the good with the bad.) The following is the progress we've made in one afternoon, a Friday afternoon. I noticed as evening came, the AI got stupider, truncating responses to save bandwidth. I paused the session to continue when it is less taxed by millions of users hunting for cheat codes for their Friday night gaming sessions and the like. I think my labor-intensive queries were forcing the algorithm to cut back my bandwidth to deal with other user's queries.
Rongorongo Translation Attempt Full Session (with Archive File at End)(.rtf)
Exigent circumstances forced me to wrap up the work, my power was about to be cut for a utility pole replacement, but I think we made really good headway with the language. There are some partial translations near the end of the second "Full Session" document and an archive file that can be inputted to continue the conversation for linguists interested in making further queries to the AI for themselves. Please give citations for Google's AI and myself if you use it for published research papers. The AI is the translator, I am merely the prompter that teased it out.