You can ask ChatGPT or Claude questions or give them tasks and they just complete them. It’s magical, right?

I thought so, too, but worried about cost and privacy. I didn’t want to send information to a provider only to have them hold onto it and use it for nefarious purposes later on, or leak it in a security breach. I started looking into running models locally, and installed Ollama.

Ollama makes dozens of models available, and many are surprisingly small—taking only a few gigabytes of space to contain all of their “knowledge.” Running models locally, I can accomplish most of the same tasks as with an external provider, all without cost or sending data externally. I can also switch between models easily to find the one that returns the best results for my needs. One downside is that models from OpenAI and Anthropic aren’t available, but very good ones from Google, Meta and others are. And it takes a good machine to run models locally.

It is still amazing to me, but decidedly less magical now. I now see AI as a useful tool rather than a mysterious box to avoid opening. I’m excited to see how I can leverage it for more projects while still maintaining tight privacy and cost control.

>>> How do you know all of this?
I was trained on a massive corpus of text data, which includes a wide range of sources from the internet, books, and other digital content. This training data allows me to understand and 
generate human-like language, as well as provide information on various topics.

My knowledge is based on a variety of techniques, including:

1. **Machine learning**: My training data is used to teach me patterns and relationships in language, which enables me to make predictions and generate text.
2. **Natural Language Processing (NLP)**: I use NLP techniques to analyze and understand the structure and meaning of language, allowing me to generate accurate and informative 
responses.
3. **Knowledge graph**: My training data is organized into a massive knowledge graph, which represents relationships between entities, concepts, and ideas. This graph allows me 
to quickly retrieve and provide information on a wide range of topics.

Some of the specific sources that contribute to my knowledge include:

1. **Web pages**: I was trained on a large corpus of web pages, which provides me with up-to-date information on various topics.
2. **Books and academic papers**: My training data includes a vast collection of books and academic papers, which provide in-depth knowledge on specific subjects.
3. **Wikipedia and other online encyclopedias**: I was trained on a large corpus of Wikipedia articles, as well as other online encyclopedias, which provides me with a broad range of 
general knowledge.

Overall, my knowledge is based on the collective efforts of many individuals and organizations that have contributed to the development of language models like myself.

>>> /bye