1.0. Introduction Generative AI or GenAI
Today, I learned something new worthy of sharing with you. I learned how artificial intelligence (AI), specifically generative AI, could revolutionize your work. Do not forget, I am not exceptional to the same impacts.
Undoubtedly, artificial intelligence is taking us by storm and rapidly changing our daily lives. It is changing how we do things right, from health, communication, and education to finances and the economy in general.
Do you use GenAI in your daily life? This is a ‘Yes or No’ answer question. I bet most of us will say ‘No’ due to unawareness. However, even if your answer is ‘No,’ it does not mean you lag; it is only that you are unaware. Again, you are not alone because I have been using it without knowing it until recently. Since I have been there, I believe everyone using a smart device uses it.
1.1. How I learned I was using Generative Artificial Intelligence (GenAI)
In my case, I use it in several ways, including learning new things, solving questions, and sometimes generating ideas. Initially, I was unaware I was using generative AI when I autocompleted my sentences in my emails on Gmail on my phone and PC. See the screenshot below. Have you come across something like this before?
Therefore, I learned I was already using GenAI when I received a marketing email from Sololearn, a learning App. Previously, I used the App to learn Python during my data science course. The email asked me to be among the first to interact with their new GenAI course.
As a result, when I started learning the course, I realized that I had been using it. The course taught me how Generative Artificial Intelligence picks the next word and suggests it to you, thus the autocomplete phenomena I mentioned earlier.
1.2. Interacting with Generative Artificial Intelligence
This comprehensive research-based tutorial will teach you how to interact with GenAI tools to transform your daily life and work. Also, you will learn and master how to use the tools to create, automate, and become more productive. Specifically, you will be introduced to GenAI, the art of prompting, Large Language Models (LLMs), and how they work.
Now that you have a solid background in the tutorial’s subject and purpose let’s explore the details.
1.3. Examples of real-life ways you can tell if you have used GenAI before
If you see any word suggestions when writing a message on your smartphone, it is GenAI. Also, if you have seen Gmail’s word suggestions when composing an email, that is Generative Artificial Intelligence at work.
Additionally, if you have used the ‘tab’ button on your keyboard to auto-complete your statements in email messages or chatbots prompts, that is artificial intelligence. For example, in the screenshot below I show you how Copilot pmopted to press “tab” to complete my statement.
Therefore, as GenAI assistant applications become popular each day, it is essential to understand how their magical human-like responses are powered and presented to you as expected.
As part of learning how to interact with and use GenAI tools to your advantage, I will explain:
- How Generative AI applications that we use daily in real-life scenarios are powered by Natural Language, a Language Model, and large language models (LLMs).
- How the LLM temperature setting controls GenAI creativity.
2.0. What is Generative AI and How is it Related to Artificial Intelligence?
Before I define each of the two concepts, here is a brief introduction to this section. As stated earleir, Generative Artificial Intelligence is abbreviated as GenAI while Artificial Intelligence is abbreviated as AI. Even though both AI and GenAI belong to the same technological field, they are not the same. I know you may quickly ask yourself but how? As a result of this question, I have a quick answer to it.
As shown in Figure 1, GenAI is considered the latest innovation in the field of AI technology. This is demonstrated by the fact that the Artificial Intellience field on the chart houses GenAI. In other words, GenAI is contained or found under AI. Thus, we consider GenAI the latest innovation in the field of Artificial Intelligence technology.
Currently, AI is revolutionizing the world. But how does it work together with regards to its subsets such as GenAI? In the next section, I will be answering the above question in details. However, before we answer it let me briefly give you a brief overview of AI. This will prepare you for GenAI.
2.0.1. What is Artificial Intelligence (AI)?
Generally, artificial intelligence (AI) refers to the science of making machines smart or intelligent like humans. Also, we can refer to it as designing and developing applications that use human intelligence to perform tasks previously said to be performed by humans only. Thus, it enables automation, decision-making, and problem-solving across industries.
2.0.2. Artificial Intelligence and Its Subsets: A Simplified Breakdown
AI is a broad category of revolutionary technology that is made up of several subsets. For us to clearly understand how AI works, we need to understand what powers it. That is the individual subsets right from the smallest to the largest one – see Figure 2.
Let us break AI down from the top downwards.
The substes of AI that power its applications are:
i. Machine Learning (ML):
It is the first subset of AI. ML is the type of AI that learns patterns from data to make predictions. Therefore, Machine Learning improves over time as it processes more data. As a result, the continuous learning process makes it essential for applications like recommendation systems and fraud detection in different industries.
ii. Deep Learning:
A subset of Machine Learning using neural networks inspired by the human brain. It powers speech recognition, image processing, and self-driving cars by analyzing vast amounts of data.
iii. Generative AI:
It is also known as GenAI. It is the Artificial Intelligence that creates new content like text, images, and music following user instructions. Generative AI works by learning from existing data and generating human-like outputs, enhancing creativity and automation.
iv. LLMs:
LLM is the acronym for Large Language Models. These are specific and advanced AI models like ChatGPT, Gemini, Claude, and DeepSeek that understand and generate human-like text. They process massive datasets and use deep learning to improve language-based tasks.
2.1. Uses and Commonly Used Types of Generative AI Tools
We can all agree that AI technology advancements, GenAI, have revolutionized how various industries work, from health systems to banking in our financial sector. This is because GenAI can used for various tasks. Most of these tools are designed with specific tasks in mind. Generative AI tools can be used to create new and unique content, including images, text, audio, and videos, based on user descriptions.
GenAI has also changed our daily lives and the work environment. For example, Grammarly.com laid off approximately 230 workers to adopt AI for a futuristic workplace.
Due to its capability to generate new and quality content, GenAI is used to;
2.1.1. Text/Content Generative AI: Used for Generating New Text or Content
With the use of an AI Chat, a user can create new text by simply chatting with the Chatbot using text messages. To make it even simpler, the messages are written in their natural language. This means that the AI Chat will understand human language and respond with the most appropriate human-like responses in a content format. For example, I used ChatGPT 3.5 and Microsoft Copilot GenAI chatbots to generate the content on the two screenshots below.
2.1.1.1. An example of how to use ChatGPT to create new content.
In the screenshot below, I used ChatGPT 3.5 to create new content using a simple prompt. First, I opened the AI tool and typed the prompt. Next, I pressed the send button and the tool responded in seconds.
2.1.1.2. Content Creation Example using the Microsoft Copilot AI Tool.
2.1.2. Image Generative Artificial Intelligence: Used for Editing or Creating New Images.
Similarly, GenAI creates new images or edits existing ones using user text descriptions. For example, a tool like DALL-E can generate new realistic photos from your text descriptions. The screenshot below shows the image generated using the Microsoft AI Designer powered by DALL-E 3.
Furthermore, you can use the Adobe Generative Fill tool to edit existing images.
2.1.3. Audio and Video Generative AI tools: Used to Generate Audio and Videos Based on User Text Descriptions.
Additionally, you can write specific text descriptions using the GenAI tools to generate audio or audiovisuals that meet your needs. Also, you can use tools to edit any of your previously created audio and videos. These audio-generating AI tools with various applications include audio generators, AI music generators, audio enhancers, text-to-speech tools, and AI video generators.
Finally, specific examples you can use to create, generate, or edit your voice include ElevenLabs, WellSaid, LANDR, Speechify, and Descript, among others. Generate everything to do, audio and videos based on your text or descriptions written in the natural language of your choice. I do what I say – here is a sample video i generated using the VEED.IO tool. You can use the same tool to create your own videos.
2.1.4. Text analysis and classification
Moreover, the GenAI tools can help you analyze and classify text. They work in conjunction with large language models (LLMs) to perform text analysis tasks for us. Are you wondering how they do that? Do not worry; I will explain later in this tutorial how natural languages and LLMs work together to analyze and classify text.
An example of a tool that we used to analyze text before AI came was Grammarly. Since it integrated with GenAI, it added text analysis functionality to generate classified content from scratch or edit existing content.
2.1.5. Code Generative AI: Used for generating program codes.
On the other hand, code-generative AI makes it possible to create programs automatically using sophisticated algorithms that understand programming languages and patterns. It is based on learning how to generate new functional code according to predetermined requirements from existing codebases.
AI systems predict and generate code pieces, functions, or entire applications by applying machine learning techniques, hence broadly improving the speed at which development is achieved. This technology harnesses vast amounts of data and computational power to continuously improve its ability to autonomously write efficient and reliable programs.
3.0. How the Generative AI Works
Generative artificial intelligence is designed and developed in a way that it understands most natural languages and conversationally replies to them. Due to this capability, we can communicate with it like a fellow human being. We give it input in different formats, such as text, image, video, or audio, and it generates an output.
As a result, we can use GenAI to work on tasks just like we would do with a workmate. Therefore, based on the above types and uses of GenAI, it works on two crucial things: natural language and language models. Later in this tutorial we shall look at how each works.
First, let us briefly look at the two terms.
3.1. What is a Natural Language?
By definition, a natural language refers to the language spoken by humans. As a result, we can communicate with AI assistants or Chatbots and receive human-like responses as we do with our fellows in our language. Therefore, we can give GenAI tasks to work on together to improve productivity. Natural language is the secret to its magical abilities to jungle millions of stories and generate real-time human-like responses.
3.2. What is a Language Model in GenAI?
Similarly, a language model refers to a specific program capable of determining which words are likely to follow each other based on context. Also, the program analyzes corpus sequences to determine which words follow. Thus, a language model powers next-word predictions by detecting which words will likely follow others. It learns from a corpus.
In simple terms, a corpus is a body of related text data on a subject. For example, if a language model learns from text data from medical records, then it is a medical corpus. Finally, as the corpus grows and encompasses huge and diverse subjects, it forms a Large Language Model (LLM).
4.0. How We Interact with Generative Artificial Intelligence (GenAI)
First, note that not all GenAI applications are conversational, and not all conversational AI applications are generative. Some examples of these conversational Artificial Intelligence assistants include Google’s Gemini, Open.ai’s ChatGPT, and Jasper AI. However, the conversation between us and any conversational GenAI assistants starts with a simple or complex prompt.
4.1. What is a Prompt?
A prompt is the statement you enter into the AI Chat interface. It is also known as the input. After you input a statement, the GenAI gives you a response based on the input. As a result, your input query determines the quality of the response received. For example, if you write a poor prompt, the AI assistant will respond inadequately.
Since the output solely depends on the input, one prompt can produce two different answers depending on how you write it. Besides, you can edit or add more details to your prompt to improve the final results. This is because the technology is flexible and thus very creative.
4.2. How does Artificial Intelligence (AI) Choose the Next Word?
The language model uses two methods to choose which word to follow. These are;
4.2.1. Probabilities.
In this case, first, the language model measures how frequently words follow each other in various contexts and the corpus sequences of words. Then, it calculates the probabilities of each word being selected.
Next, the probabilities are then represented as percentages for better understanding. For example, a word with a probability of 51% is more likely to be chosen than 13%.
After the word is chosen and put in the text, the language model repeats the same process to select the next word until the output is complete. As a result, the final production can be an email, a sentence, a paragraph, or several paragraphs.
4.2.2. The Temperature Setting.
Secondly, the temperature feature is another crucial element of a language model. It controls and influences the randomness or creativity of the model. This means that it also influences the final solution or output of an AI assistant model. At the lowest temperature, the model is said to be sleeping, and there is no randomness.
In other words, the input is similar to the output. Therefore, as temperature increases, model randomness increases. At very high temperatures, uncommon and unlikely words become probable choices or likely.
For example, “Playing” is the next word because the probability is 100% at the sleeping level.
If we adjust the temperature by sliding the button to the middle level on the bar, it is 51% likely to be selected.
4.2.2.1. Results of adjusting the temparature on the bar.
As a result, other words start to become likely because their probability increases. We can verify this by looking at the new probability of other words. The words “eating” and “sleeping” gained probability of 13% and 16% respectively. In addition, the remaining percentage was allocated to “driving” (9%) and “flying” (3%). Summing up the various probabilities, you will get a total of 100%.
Furthermore, when the temperature is highest, all the words have the same probability of 20%. This means that AI can select any of the words as the next one.
If you calculate the total by multiplying the percentage probability by 5 words (5 x 20%) you get 100%. This means that the language model is less creative and more random at very high temperatures. At this level any word can be chosen and added to your sentences or paragraph thus generating a less accurate output or content.
5.0. What are Large Language Models (LLMs) and how do they work?
These are language models whose corpus contains massive internet-based text data on various subjects. The text data include books, public conversations, articles, and webpages. The text data is in different languages. Based on this, most Large Language Models are multilingual.
Therefore, they are powerful natural language experts who are good at predicting the next word or text completion. They give GenAI the ability to understand and generate natural language. Thus, they communicate effectively in two-way chats with humans and at a high degree of accuracy.
5.1. How Large Language Models Work in Generative AI.
This is where you learn what tasks you can give to LLMs and leverage their power to be productive in your job. They are good at following human instructions. Since the AI Assistant or Chat is connected to LLMs, we work with them as teammates. Besides, we give them tasks to work on, saving ourselves time because they can follow our instructions. In summary, since LLMs are excellent experts at following human instructions, they can be used for the following.
- Building Chatbots and Artificial Intelligence (AI) assistants: relies on understanding and processing natural language.
- Content Classifications: use its analysis and categorization abilities.
- Writing and Reading: Use them to generate new content or read other texts.
- Automation of daily and repetitive workflows: Create scripts to automate your workflows for efficiency.
In Summary:
The article has extensively introduced you to Generative Artificial Intelligence (AI), popularly known as GenAI. You have learned how Generative AI applications are powered using natural language, a language model, and the Large Language Models (LLMs). Additionally, it has taught you the types and daily uses of GenAI tools. They include text or content generation, audio generation, creating videos, and editing or creating new images.
Furthermore, the tutorial described and demonstrated how AI chooses the next word and adds it to the text. We learned that it uses probabilities represented as percentages and the temperature setting for a language model.
Additionally, it has explained how LLMs work and how we use them in our daily lives. Such uses include Chatbots, Classifications, Writing and Reading, and Automation. As a result, now you know how to influence the creativity of any GenAI application you use.
Similarly, you can tell when the responses you receive from your AI Assistants and Chatbots are likely or unlikely to be accurate. This is because you know that the temperature setting of the language model influences responses by influencing the next word selected and added to the text in a repeated manner. Therefore, go out and apply the knowledge gained in an excellent course. Finally, if you like our content, refer your friends we learn together and subscribe to our newsletter. To learn more about AI you can visit our AI category.
