Paleo.In

Status

This post explains what we are currently doing in order to try and find the inspired text of the Bible. Read on for more.

Between Letters and Words

For several years I have believed the key to finding the inspired text has been in the gap between individual letters and how those letters are arranged in written documents as running text.

People do not write with a letter budget. Think about how people write prose. Good writers start with an idea they want to convey to their readers. From that sort of starting point they build up the story they want to tell, eventually writing down a series of sentences.

This is obviously a different task if you are writing a novel, versus writing a short email, but the principles are the same.

Once writing sentences, for regular humans, there is no restriction on available letters. Writers can use whatever letters, in whatever quantity, they need to express their story.

There are almost no writing tasks where there is some sort of 'budget' that constrains available letters. An interesting example where there is a letter budget are crossword puzzles.

Crosswords

Think about writing a crossword puzzle. The most constrained example I am familiar with is a typical NYT puzzle. The constraints are these: 1) The puzzle is square. 2) All black squares (punctuation) are symmetric around the center. 3) All letters are part of both an across clue and a down clue. There are other constraints based on the day of the week, such as panel size and clue difficulty.

Building these sorts of puzzles is an art form. Here is a Video of an actual NYT puzzle setter showing how it is done.

Puzzles like this are usually focused on a theme. The clues for the words in the puzzle come from some intellectual domain. Say sports, or farming, or cars, or something similar. In the video above it is playing cards.

What you do not find is all the across words forming complete proper sentences and at the same time all down words also forming complete proper sentences. Why?

That is a constraint that humans cannot satisfy. Perhaps an AI system on a computer somewhere might, but not regular humans. Especially not humans from history as we know the text of the Bible is from history.

We do not think with a letter budget like this when we are writing stories. We do not write more than 1 story at the same time.

I have been using this example when I speak to people about the project of textual recovery. There was some letter budget, or letter level constraint, that no person could have ever met, but which the 'mind of god' could satisfy.

I have looked at other letter level constraint systems in previous years, looking at localized constraints, where letters were constrained by nearby letters. These mostly worked at eliminating added vowels, but no more.

It did not dawn on me until early this year that this example of a crossword was probably very close to the actual constraint system used to write the text.

This is especially interesting if you think of being passed, say, only the across version of the story. Anyone, from say, Junior High forward, who knew the inspired language, could check inspiration by arranging the letters on a grid and then reading off the other document.

This does not require any wizardry for a proof of inspiration. All the other systems I've studied over the past 10 years required code wizardry for confirmation of inspiration. Not this one.

There are parables about this too. The document passed by history is called Adam. Eve is the second document that springs from Adam's DNA. The second monument, the second scrolls in the vault, and so on, are all Eve documents.

What I have been working on for the past few months is trying to prove this crossword style inspired form is what was going on with the original inspired text.

Of course this is not in exactly all the details used in NYT crosswords, but something very similar. Every detail must be deduced in some clever way. Very similar to the way the alphabet's letter designs have clever design constraints.

I am pretty convinced of the panel size, and of a boustrophedon letter arrangement. What I don't have is a full grasp of the inspired language itself. We are not passed Adam clean, we are passed Adam severely edited. Both word replacements and spelling changes have happened across the centuries.

So the real problem is making educated guesses on Adam and looking for Eve via computer code. But those guesses must be made across an entire panel of letters, not individually nor strictly with neighboring letters. This is a massive computer problem.

Stats

When I last blogged about this I was running about 20,000,000 test cases and it took about an hour. This was interesting, but way too slow. Realistic test cases are more like 20,000,000,000,000. I am rounding down, but this is the count for cases needed to study the first panel in the commandments. As you can imagine, I don't have ready access to 1,000,000 hours of computer time.

So, what to do in this situation? I normally just go do something else while hoping for inspiration.

I have been putting a web front on the code, so that test case prep is a matter of clicking around and pushing a few buttons. I can now go anywhere in the text and get a default test plan for any verse. And, I can send that plan to the solver.

Provided, of course, the text has been properly prepared. Once we know what inspired text looks like, this will be easier to do and will reduce test cases. But, right now, nothing is known with certainty, everything must be tested.

I have also been working the solver itself. Out of Javascript into compiled C, and ready for CUDA C on the GPU.

In the past week I was hit prophetically with the solver routines themselves. First by breaking it down into stages, and then by a strange recursive rebuild of the code itself that came to me in a flash one morning before breakfast this week.

After some fiddling around I am now running those full 20 trillion cases needed to understand the commandments in around an hour. It has been a wonderful week.

Maybe around 1000 of those generate legal, interesting, Eve documents. Each of them is worthy of further study.

Still missing much display code for looking closely at candidate Eve texts. I will be turning to that code next week.

Please keep praying for the work. This is mostly a prophetic journey, not so much a technical problem.

More Later,

Phil