Artificial intelligence (AI) is one of the fastest emerging technologies today. Several AI tools and platforms use different AI models and various data sets for training and fine-tuning algorithms to achieve the best results.
Understanding AI modeling can help you choose the best AI tools based on your needs. Learning how AI models work can also help you write more effective AI prompts that lead to better outputs.
This guide will cover the different AI models and outline how they work. You can also try Atlassian Intelligence or Rovo for free to see our AI solutions in action.
What is an AI model?
An AI model is an artificial intelligence computer that uses algorithms to learn from datasets, make predictions, generate content, or produce other outputs. Developers use different datasets and algorithms to create AI models specializing in various tasks.
While some AI programs use the same underlying model, there are key differences among the various tools and programs. Understanding these differences is essential to choosing the right AI model for your needs.
Types of AI models
There are three primary types of AI models: supervised learning, unsupervised learning, and reinforcement learning.
- In supervised learning, each data set has predefined labels that allow for more specific training. Several algorithms are used in supervised learning models, which are commonly used for classification and regression.
- Unsupervised learning starts with data with no predefined labels, meaning the AI model is responsible for identifying patterns within the data. This can be useful for segmentation and association.
- Reinforcement learning involves more than just analyzing a data set; the AI model receives feedback to reinforce positive actions and discourage negative ones. These AI models are often used for automation in the robotics and self-driving car industries.
Comparing AI, DL, ML Models, and Agentic AI
You’ve probably heard terms like “deep learning” and “machine learning” used interchangeably with AI, but they mean very different things. This section explores alternative models that achieve similar outcomes through different approaches.
AI models
AI is an all-encompassing term that includes various types of models, such as deep learning and machine learning. AI models use artificial intelligence — which consists of a wide range of capabilities — to automate tasks, solve problems, and generate content. Generative AI models allow you to create music, videos, and images from a simple text prompt.
For example, AI in Atlassian allows you to automate repetitive tasks, generate content summaries, and refine existing content.
Deep learning (DL) models
Deep learning (DL) uses multi-layered neural networks to emulate the human brain. Each layer processes distinct pieces of information and learns from the input before transmitting the compiled data to the output layer.
DL models are typically used in applications like self-driving cars, facial recognition, and natural language processing (NLP).
Machine learning (ML) models
Machine learning (ML) models are similar to deep learning (DL) models, but they’re less complex. ML models analyze datasets to find patterns, which they learn from and use to make predictions. Deep learning requires more data and can achieve better results with certain tasks.
Agentic AI
Agentic AI is an artificial intelligence system that goes beyond simply responding to prompts or following pre-programmed instructions. It’s designed to be more autonomous, proactive, and adaptable, much like an intelligent agent.
Think of agentic AI as an agent in the form of AI. If you hire an assistant, you expect your assistant to change and adapt as your needs change. Agentic AI can adapt to changing situations, using patterns to make decisions. Agentic AI has a memory, which allows it to store and learn from past experiences and use that previous data to inform future decisions and actions.
For example, an agentic AI-powered personal assistant could learn your scheduling preferences over time, automatically prioritizing meetings based on past choices and adjusting reminders accordingly.
How do AI models work?
AI helps your team work smarter, but how do the underlying AI models actually work? Let’s take a closer look at how AI models are developed using data sets and how AI tools use your input to generate content, make predictions, and take action.
Every AI model starts with large amounts of data. AI models learn from the data you feed them, identifying patterns and relationships within a data set. In some cases, each piece of data is labeled to guide the AI training process — this is known as supervised learning. Other models use unlabeled data that allows the AI model to explore patterns and relationships and draw its own conclusions using a set of analytics.
In most cases, AI models are supervised and adjusted during the training period to achieve the best results. By making slight adjustments to an AI model during the training stage, you can change the algorithm to generate the results you’re looking for.
AI modeling can also change when models have access to new data. For example, ChatGPT was only trained on data up to September 2021 when it was first released. Now, ChatGPT has been trained on newer data which allows it to provide up-to-date information about recent events.
Applications of AI models
AI modeling enhances efficiency and productivity in many organizations. Various industries can benefit from integrating AI models into their workflows. For example, project managers use AI to automate repetitive tasks, organize projects, and allocate resources based on project needs.
While AI was initially used to automate simple tasks and generate basic content, it’s becoming increasingly common in software development. AI allows software developers to streamline coding, testing, optimization, and deployment.
One of the most common applications of AI today is customer service. You’ve probably noticed that most of the largest companies use AI chatbots to answer simple questions and redirect customers to live agents. Using AI to simplify and optimize customer service can improve response times and enhance customer satisfaction.
Building and training AI models
Building and training AI models involves four main steps: data collection, model selection, training, and deployment.
- Data collection: AI modeling starts with collecting data you can use to train your AI model. The data you use significantly impacts what your AI model learns and what it’s capable of. Properly processing data is crucial, which involves ensuring it is accurate, clean, and consistently formatted for effective learning.
- Model selection: Next, you have to select an AI model type. Supervised, unsupervised, and reinforcement learning all have strengths and weaknesses, so the focus here is choosing a model type that can do what you need it to do.
- Training: Once data is collected and the model type is selected, use the data to train your AI model. The training process will vary slightly depending on the type of model you’re building. Feedback loops play a key role in reinforcement training by helping to guide the model.
- Deployment: Now, it’s time to deploy your AI model. You can host your AI model on a website or deploy it on your local system. Even after deployment, you can continue to improve the model by adding new data and refining the training process to enhance its performance.
Examples of AI models
There are many successful AI models, including GPT-4.5, Atlassian Intelligence, and Rovo. Different AI models excel at different things—some are commonly used to generate images, while others are better for automation and software development.
Atlassian Intelligence
Atlassian Intelligence is a comprehensive AI model that integrates with other Atlassian tools and services. Atlassian Intelligence can enhance collaboration when used with Confluence whiteboards, create AI summaries of content, and generate and refine content.
Rovo
Rovo is another popular AI tool that enhances productivity and streamlines projects. You can start with our pre-built Rovo Agents or build your own. You can even design each agent to perform specific tasks or integrate with your existing software and services.
Unlock the power of AI with Atlassian Intelligence & Rovo
AI has become one of the most powerful technologies in recent years. However, with the growing number of AI models, choosing the right one can be difficult. At Atlassian, our commitment to responsible technology principles ensures that we prioritize trust in the development of our AI models, empowering everyone along the way.
If you’re new to AI, getting started with Atlassian is as simple as it gets. Try Rovo free to see how Rovo Agents can streamline workflows and enhance productivity, or try Atlassian Intelligence to seamlessly integrate AI capabilities with your favorite Atlassian tools.