AI For The Rest Of Us
This is placeholder text. To connect this element to content from your collection, select the element and click Connect to Data.
RushTree, Large Language Models & Prompt Engineering: Volume I/Issue V
February 26th, 2023
Thank you for your messages, emails and virtual meeting requests asking about RushTree Business Optimization and what inspired us to launch an Artificial Intelligence (AI) focused agency years before many were paying attention to AI or even heard about ChatGPT.
The vision for this newsletter and for founding RushTree Business Optimization is to share knowledge, AI tools and insight in a manner that is easy for business owners, managers, and busy professionals to understand and use... All at the speed of NOW.
What is the speed of NOW? Well, the average attention span of humans, (8 seconds) is less than a goldfish, (9 seconds).  Our average attention on screen has dropped from 2.5 minutes in 2004, to currently an average of 47 seconds. In addition, once we are distracted from an active task, it takes about 25 minutes to refocus on that task, . Finally, some would argue that professional etiquette and authentic team/company culture is almost non existent. We need a better way forward in work and in life.
RushTree has the solution. We are industry and role agnostic. We deliver easy to implement, universal frameworks to optimize all aspects of your business at the speed of NOW. We combine time tested models based on 6-Sigma Total Quality Management, (Simplified) with proven, suggested tools (AI Software & Hardware) that makes it easy for busy people and teams to execute on assigned projects and tasks, with minimal disruption to their workday. RushTree solutions optimize performance, profit and culture. This is essential to thrive in todays' challenging and dynamic work environment.
Why focus on AI? With any complex construction project, you need a blueprint and a tool set. RushTree solutions provide simplified blueprints and identifies where and how to use the right artificial intelligence, (AI) as your toolkit. Think of using a handsaw to cut one log during the course of a day. No problem. But, if you need to cut 100 longs by the end of the day, to reach your goal, you'll need a chain saw. AI is the chain saw, along with an entire new set of power tools. It would behoove all of us to master the language, logic and logistics of AI. We must adapt to stay relevant and succeed at the speed of NOW.
RushTree offers 3 categories of solutions branded as PULSE, MAX & THRIVE. More can be learned by visiting us at rushtree.com. RushTree Business Optimization is your partner to succeed at the speed of NOW.
LARGE LANGUAGE MODELS (LLM)
"LLMs are usually very large (billions to hundreds of billions of parameters) deep-neural-networks, which are trained by going through billions of pages of material in a particular language, while attempting to execute a specific task such as predicting the next word(s) or sentences. As a result, these networks are sensitive to contextual relationships between the elements of that language (words, phrases, etc).
For example, "I was sitting on a bank of snow waiting for her". What is the meaning of "bank"? Bank - an institution, bank - the act of banking, a riverbank, or a plane banking to the left or right, or any other? While it is an easy task even for a child, it is a nightmare for a computer.
Previous models have been stuck at 65% accuracy for decades, but now a standard BERT based (LLM) model is able to do this in a reasonable time (milliseconds) with an 85% - 90% accuracy.
As they are tweaked and improved, we will start seeing a shift from using AI for static tasks like classification, which can only serve a small number of use cases, to entire linguistic processes being aided by machine learning models, which can serve a tremendous amount of use cases.
We already see such applications through ChatGPT, Github Copilot, and many others." 
Image source page on Medium.com
One key to optimizing Large Language Models, (LLM) is to master the skill of prompt engineering. The better the input, the better the output. When developers are fine tuning or training a LLM, you may hear them use prompting terms like:
Role Prompting, [respond as a world-renown scientist with an expertise in neuroscience]
Shot Prompting: (Zero shot , One shot, Few Shot, etc.) A Zero Shot Prompt example is a direct question: What is the capital of North Carolina? A One Shot Prompt example would typically include a task description, along with an example: List the actors and their role in the movie, "Rocky": [1. Sylvester Stallone, "Rocky Balboa"]
Chain Of Thought Prompting: When we encourage the LLM to show its reasoning or computing in its' response. This works well for computational and reasoning problems. For example: What is 3x1? [1+1+1=3].
As AI power tool users, we will focus more on key tips to optimize our prompting commands in applications that access Large Language Models, for example, ChatGPT.
Here are 5 key tips to help you optimize prompt engineering:
Be Clear and Specific: When entering your prompt, use clear and specific language that the model can understand. For example, instead of asking "What's the weather like?" you could ask "What is the current temperature in New York City?"
Provide Context: Providing context and details is important to help the AI model understand the intent behind the prompt. For example, instead of requesting "Create a blog post on Dentistry" you could provide context by requesting, "As the world's top advisor to Dental Practices, create a 900 word blog post on the advantages of utilizing AI in the Dental Practice. Include examples for both the clinical and administrative aspects of the practice."
Use Examples: Providing examples can help the AI model better understand what you're looking for. For example, if you're asking the model to generate a list of programming languages, you could provide examples such as Python, Java, and C++.
Test Your Prompts: Before using your prompts, it's important to test them with the AI model to ensure that they're generating the desired output. For example, you could test a prompt by asking the AI tool to generate a sentence about a specific topic and evaluating the accuracy and coherence of the output. If the output is weak, ask the tool to regenerate the response until you acheive the desired output.
Continuously Refine: Refining your prompts based on the output you receive from the AI tool can help improve the quality of the output over time. For example, if you notice that the model consistently generates inaccurate or irrelevant responses, you could refine your prompts to provide more specific context, details or examples.
A quick Google/Youtube search or asking ChatGPT to provide examples of prompting best practices should supply some good direction and inspiration to develop your prompting skills.
AI beyond ChatGPT: You are going to interview 2 AI chat tools!
Follow the steps below:
Visit each AI chat tool, by clicking on their images below in this article.
Enter the following questions and read the response. (You may need to create follow up questions to get desired results).
Who are you?
What Large Language Model(s) do you use?
What are your strengths?
What are your weaknesses?
What is, [AI chat tool name] purpose in life?
3. Bookmark these sites, add their chrome extensions and share them with family, friends and colleagues to advance, "AI for the rest of us"!