AI - #2 (Sample LLM)
NOTE: Within the storm and complexities of Chat GPT, Bard, etc. etc. I am attempting to keep it simple and fun, yet AI is often neither. Simple is relative, but hopefully even technical folks will enjoy a perhaps unique look AI. Also, forgive the sometimes mocking tone, it’s entirely intentional. I will make futile attempts at humor, satire, sprinkled with sarcasm, but AI deserves it. RIIIGGGHHHT. Bear with me and enjoy the topic!
To start, I will quote the famous football coach Vince Lombardi “This is a football”. Well, in Artificial Intelligence or “AI” from now on, “This is a computer”.
The basics of AI include Neural Networks, Large Databases (the silicon crowd will say “massive”), and lot’s (technical jargon meaning more than you have) of Compute power. We will work on defining each of these in exec level detail as part of this series. Use of the large databases can be simplified to an acronym (of course, did your Mom not tell you to never use TLA’s?) called LLM’s (Large Language Models). We will expose LLMs (I mean explain) these in more detail as part of the series.
How do I know any of this to be true? Will I looked it up on ChatGPT and Bard of course. Type some questions into these engines (or similar) and you will get answers along these lines, so it must be fact.
So what are some intended or theoretical uses for “AI” and these new models ?
Generating text such as poems, code, scripts, musical pieces, email, letters, bad rap lyrics, and most importantly, plenty of fake news reports. Say bye-bye to all the talking head news models and “journalists” writing for the New York anything.
Translating languages such as English to Spanish or French to German. Most importantly for our future however, is Chinese to English.
Creative writing seems counter intuitive to an AI system, but the claim is there. Blog posts, articles, poems, stories, scripts, musical pieces, etc. are generated from thin air (or LLMs) and will make grading papers for what’s left of academia a near impossibility … unless of course the teachers grade the papers with AI (touche!). Wait … we don’t need students or teachers anymore! Problem solved.
Ok, before we jump off into the deep end, we can start generating some pseudo code of how and AI LLM model/engine might work in the “real world”.
Intentionally Simplified LLM functional model to demonstrate and outline the pseudo code sample below.
Below is a simple example of an AI LMM in pseudo code:
// **** Psuedo Code AI LLM example
function LLM (input_data, output_data)
// Initialize the neural network
network = create_neural_network(input_data, output_data)
// Train the neural network
train_network(network, input_data, output_data)
// Use the neural network to generate output data
output = generate_output_data(network, input_data)
// Return the output data
return output
end function
Basically, what is happening in the above “code” is the system is inputting a neural network model, some context related and very large data source(s), then combining this information and creating a “trained” LLM model that can then be interrogated and interpreted for use by humans (or other AI models or systems). These model core components, neural networks, databases, training, and compute aspects of these AI LLM’s will be deconstructed and explored in coming segments as part of this series.
AI practice and Task of the Day: Use Bard or ChatGPT (or others) to work on 3 tasks per day. See if it is useful tool. For example, I was asking Bard each day for the biblical verse of the day. The first couple of days were great, with a verse and useful explanation of the verse’s meaning. However, on the 3rd or 4th day of this practice, it began repeating itself. Hmmm, that was not so good. Even changing up the questions was not overly helpful in avoiding the repetition. I will keep trying. I am sure to expect ever improvement of these products as the develop from version 1.uh-oh and beyond.