You need to know the answer first ....

Message:


The way I saw it was you feed it a "small" problem that you know the answer.

And, the ASS adjusts the weights.

Then, you feed it another with the answer and adjust the weights.

Repeat. Repeat. Repeat.

When you reach the point that the weighted results matches the answer you expect, you have the right weights.

Then, feed it a problem where you don't know the answer.

Hilarity results.

In an early part, he determined the word pairs were important. If you use "Bank" and "Money", it knows (from previous analysis) those words are related.

So, the part he skipped, which would not affect his model was the syntax analysis. Take your question apart and depending on how you asked it, the model would be selected. Break the big problem into a lot of "small" problems. The tools may have already determined the answer to the "small" problems.

The code was a nerd distraction. It must somehow be in the code. Same with the blinking lights and the input hardware. Shiny.

That's why it needs the whole of something like "X" ... a big pile of confusing human nonsense. If you analyzed it log enough, you would know some times "Bad is Baaaaaaaddd".

The machine needs the data because it is ASS. The data is confusing.

Once the ASS determines the weights, it can't determine if it's wrong. It just knows the weights.

Don't give up your Man Portable Plasma Gun just yet. You may still have a need.


Yours,