Learnings on the Journey of GenAI Adoption

Learning something new is always exciting but learning something that is shaking the entire world today is way more…Gen AI, who would not have heard of the term today? Perhaps everyone presents in the digital space. But how much it is do we understand. I am going to start with the basics and then overlay my learning journey to reveal what all I discovered so far…

What is Generative AI?

To an IT professional, it is a system to generate content based on a trained dataset.

For a businessperson, it is a tool that can be used to automate routine tasks.

A creator sees it as an ecosystem to generate unique content.

If we have to sum it up for the understanding of everyone, can we say it’s any tool that can generate content that we ask for?

Content can be text, audio, video, image, code…

When we ask it…we make prompts and from that came the term prompt engineering which simply means we can write prompts in structured manner to make the AI bring out the best content possible.

Understanding the Big Word: Artificial intelligence

Now, let’s talk about the big word AI or Artificial Intelligence. Is it only about prompting? Or prompting is just a part of it or can we say a start of it?

Prompting is simply the way for us to tell the AI what we want it to do.

This prompt then goes to a system, we call the system ‘Large Language Model’ or LLM.

What is LLM? It is a machine learning model that can understand human language and can also write in human language.

Why is it called a machine learning model? Machine Learning (ML) is an intelligent system that can analyse data without getting an explicit instruction. For example, If I need it to find the area of a circle with a radius 5 cm, I do not have to feed it with the formula like we used with earlier coding languages –

…Give it the formula  where A is the area,

The machine already has that formula because it has done it at least once, so it has been trained on it, just like us humans. When we first learnt the formula and how to use it from a teacher, the teacher would then just us problems with radius value and then, we can do it on our own without help. Similar is the case with a machine learning model. Once it has been guided to perform a task, it can do it on its own. This is what we call training a model on a data set.

Within the spectrum of AI, there are more words you might have heard of - Prompt Engineering, Meta Prompting, GPTs, AI Agents, Deep Learning, and more. What are these?

Prompt Engineering (PE) : When we use structured prompting to make the LLM give us the best output, we can call it PE.

Meta Prompting: When instead of creating perfect prompts ourselves, we ask a GPT to create a prompt for us so we can use it to get good content out of LLM, we are creating meta prompts.

GPT: Generative Pre-trained Transformer is simply a family of LLMs.

LLMs can fetch your data and present content in the way you want but they still work based on what you ask. They cannot take decisions on what to do or what is right. For that we go up a level with deep learning.

Deep Learning: It is an ML tool that can teach a machine to process data and take decisions on its own.

But what if you want the machine to perform much more than that? Like not just understand what we are saying but also collect data based on it, make decisions as well as perform some tasks based on it.

For example, if you ask it to respond to an email from a customer that asks about the status of product delivery (Understand us), the AI collects information about the product from your ecommerce system (Collect Data), pick the relevant data that customer needs like location or expected time of delivery (Decision) and reply to the email with the status data (Action).

If humans did it, we need a person to perform these tasks. If AI did it we need an agent.

AI Agents: It is a software that can receive instructions from us or the environment, collect data, make decisions, and take actions.

And these AI agents can be trained to take our instructions not just in the form of text but also voice. Remember AI tools like Alexa and Siri?

Evolving with AI

On the learning journey, we begin with simple prompts to interact with LLMs

…then we learn PE to do it better

…Meta Prompting to make LLMs do it

We explore the whole family of LLMs – GPTs

We create our own GPTs to be focused and fast

Then we go deep with deep learning to make AI take decisions

We create agents to perform tasks for us

And as we keep going up the level of capabilities in the AI space, we can create…

AI assistants: They can not just perform a few related tasks as agents but can manage entire workflows and might involve multiple agents.

Multimodal AI: We are not restricted to one mode of instructions anymore like Text for ChatGPT, Voice for AI Agent but we can work with multiple media types like text, voice, images, and videos.

Explainable AI: So far AI could understand us, collect data, take decisions, and perform tasks with Explainable AI, it can also explain why a decision was taken. Lets say an AI can collect patient data, detect cancerous cells, decide which type of cancer it is and then explain how it arrived at this conclusion.

Self-Evolving AI: Having done all that, now we can also have an AI that continues to learn and improve itself. It just keeps getting better and better until it reaches true human potential.

And this is the kind of AI that some people are afraid of and can lose their jobs to.

How can we be a part of this system?

Video made with LTX Studio

To start with, I am guessing if you are reading this article, you are already a part of it and has been using at least one of these tools to throw prompts to fetch content – ChatGPT, ClaudeAI, Gemini, Co-Pilot, Perplexity…

You could be a…

User: You use simple prompting technique to fetch simple content for blog articles, posts, emails, etc.

Engineer: You use prompt engineering techniques like one shot prompts, few shot prompts, chain of thoughts prompting, role assignment prompting, instruction-based prompting, iterative prompting, tree of thought prompting and so on. Each has its own perks based on what kind of content you want.

Creator: You explore not just one GPT but more of them to generate special images and refined content. You might also be learning prompt engineering deeply and creating your own custom GPTs to simplify or speed up your routine tasks with LLM.

In past few months of my learning journey, I started with writing prompts, not finding them very effective, I learnt prompt engineering and out of curiosity and to make my work life easier, I jumped into creation of GPTs (Created four of them – Competency based JD Builder, Learning Path Pro, BARS Assessment Designer, and Science Experiment Explorer)

Developer: This is where you start taking a deep dive into it. You not just create GPTs but you create AI agents, assistants, and beyond.

And this week I am entering the developer zone. No, I am not going to learn programming but I am going to try Low Code-No Code systems like Google’s Vertex AI and Microsoft’s AutoGen.

Let's see where it takes me next on this journey...

Would you like to join this jouney and learn more? Connect with me on PM Pooja Dubey

#gpt #ai #promptengineering #artificialintelligence #learnai #talmondconsulting #facilitatorpoojadubey