The Answer To Everything

Released: May 9, 2026 8:00 AM CST

Here is the forty-second episode of Quantum Foam, The Answer To Everything. It has been a little while since I last recorded and now I am trying out a new Blue Yeti USB condenser microphone this decimal of the show. The answer to the ultimate question of life, the universe, and everything is equal to 42. This is from The Hitchhikers Guide To The Galaxy. They had asked what the answer was of the summation of everything. The answer they got back was the digit 42. Automated chatting has a long history with the advent of computers. Programmers have wanted a computer to emulate human intelligence. Artificial Intelligence is either a set of scripted programs, a neural network, or some other kind of intelligence basis. There was a service back in the day where you could ask a question and receive a full sentence response from the inquiry. This was meant to be as though you were talking to somebody. What is a good vacation in the Black Hills of South Dakota? Something would come up about Mount Rushmore or Crazy Horse. These days, we have Siri, Alexa, Co-Pilot, Gemini, and there are lots of other ones. We are at the point where we need to develop cognitive A.I. It comes up every decimal of the show. There is pretty good A. I. now out of the box. Some of this stuff can be sold as a product or service. These programs allow you to A. I. generate different media content. They are a solution designed by Artificial Intelligence. We want to focus on Large Language Models. The Ask Jeeves service ended around 2010. Before chatGPT there was Ask Jeeves. It had its own concierge butler. This technology provided information in snippets. It made it easier for people to do searches.

What exactly is a transformer? I came up with a similar way to record information. Our solution was transformers. Unidirectional functions only have an understanding of what came before the target text and not after. Large Language Models include scaled dot product attention, a multi-head self-attention layer, the feed-forward functionality of the positional encoding, and multi-head attention. This is the logic diagram that chatGPT follows. Those are the labels of the functions included. In later editions of chatGPT, multi-modality was added. This allows the cross-interpretation of different types of media. This system uses strings of text called tokens. A token is about 3 quarters of a word. This type of A. I. needs to be trained on data input. There is also the embedding of vectors. This is numerical information involved in the tokenization and other parts of the system. It makes it easier for the computer to read words. The final step is that of the transformers. These are matrixes that are made by a vector database. I thought of a similar system to be used with game Artificial Intelligence. Mapping the game A. I. tendencies into matrixes is similar to a vector tokenization matrix for an LLM. We are manipulating vector databases. The logic is weighed against how much each token contributes to sentence structure. It goes from an input to an output matrix. This converts into natural language. The output matrix is a response to an input matrix given in text form by a human. Specifics are encoded using algorithms during the training process. Vectors transform from one thing to another. This is how LLMs work. There's not much more on the subject. The A. I. characters on the show are made of only ones and zeros. The Turing Test was just recently passed.