Artificial Intelligence Size-Up

Image obtained from CoPilot.

A Generative AI or GenAI like ChatGPT uses large data sets and recognizes patterns. Over around the last 10-15 years these sets have become larger and larger, and are called Large Language Models(LLM).

The user sends a prompt and the system scubs the data and provides an answer. I had great intentions on being able to describe this to you in great detail, but came to the realization that I could cause you confusion in the process. Since I feel that it is important to understand how Generative AI works I have several options to help.

First option. I straight up aske ChatGPT to tell me how it works. Here is what it told me:

AI Size-Up: Can you explain how you work for 35 to 45 year old parents?

ChatGPT:

Absolutely! I’m an AI language model designed to understand and generate human-like text based on the input I receive. For 35 to 45-year-old parents, I can assist with a wide range of tasks, from offering parenting advice and tips to providing information on various topics, helping with decision-making, or even just being a friendly conversation partner. Whether you need help with managing schedules, finding activities for your kids, or seeking advice on navigating parenthood, I’m here to lend a hand!

AI Size-Up: How do you work on a technical level?

ChatGPT:

On a technical level, I operate through a process called deep learning, specifically utilizing a model architecture called GPT (Generative Pre-trained Transformer). Here’s a simplified breakdown:

1. Training: I’ve been trained on vast amounts of text data from the internet, books, articles, and more. During this training, I’ve learned to understand patterns in language.
2. Input Processing: When you type a message, it’s converted into numerical data that I can understand. This input is then processed through multiple layers of neural networks.
3. Understanding Context: I analyze the input along with its context to generate a relevant response. This involves understanding the meaning of words, grammar, and the overall structure of the conversation.
4. Response Generation: Based on the input and context, I generate a response by predicting the most probable next sequence of words. This is done using probability distributions learned during training.
5. Output: Finally, I convert the numerical data representing the response back into human-readable text, which you see as my response.

Overall, my ability to understand and generate text comes from this complex interplay of neural networks and statistical language modeling.

Second Option: Here is a short video explaining how ChatGPT works:

https://youtu.be/4qGrteTY1EM?si=1MJZA8ud7iGX46wj

Third Option: If you want a deeper dive I recommend Cal Newport’s discussion here. He does a great job of explaining how GenAI works and what it’s limitations are.