Blog  |  October 20, 2023

Coming To “Terms” with AI, Using AI

Last time, we began our look at the “nuts and bolts” of artificial intelligence (AI) for legal professionals with a look at how AI is disrupting the business world and legal industry.

Before we can really begin to understand AI, it’s important to define several of the common terms associated with it. There are a variety of sources of definitions of AI terms out there and each of them are a little bit different. Which one is the best? That’s open to debate.

18 Important AI Terms as Defined by GPT-4

So, we decided to identify the definitions for those terms in a unique way – we asked ChatGPT (more specifically) GPT-4 to define them! We asked it for definitions of the most popular AI-related terms, which are below.

One of the things we quickly learned about using GPT-4 to provide definitions for the terms is that it will define many of these terms expansively unless you tell it otherwise. For example, the definition it initially provided for generative AI included an opening paragraph, a detailed list of seven notable aspects and examples of using generative AI, and a conclusion paragraph! That’s a bit much! So, we asked GPT-4 to give us a summary one paragraph definition of each term.

There are also countless terms that relate to AI, so we picked 18 of the most important terms that (we believe) legal professionals should know today. We’ve arranged them in an order where each should be defined before it’s referenced in another term.

Artificial intelligence (AI) refers to the simulation of human intelligence in machines, enabling them to perform tasks that typically require human cognition. This encompasses abilities such as learning, reasoning, problem-solving, perception, and language understanding. Through algorithms and computational models, AI systems can process vast amounts of data, adapt to new inputs, and execute tasks autonomously or augment human capabilities.

Augmented intelligence emphasizes the enhancement of human decision-making and capabilities through the collaboration between humans and artificial intelligence systems. Rather than replacing human intelligence, augmented intelligence focuses on AI’s supportive role, amplifying human skills, assisting in complex tasks, and enriching our understanding. The approach prioritizes a synergistic relationship, where AI complements human strengths and compensates for our limitations.

An algorithm is a step-by-step set of instructions or procedures designed to perform a specific task or solve a particular problem. Comparable to a recipe, it provides a clear and finite sequence of operations that, when followed, yields a solution or desired outcome. Algorithms can be executed by computers, but they are fundamentally logic structures and can be applied in various domains, from mathematics and data analysis to everyday problem-solving. In computing, they form the backbone of programs, enabling machines to transform inputs into meaningful outputs.

Machine learning is a subset of artificial intelligence that enables computers to improve their performance on tasks by processing data and learning from it without being explicitly programmed. It involves algorithms that find patterns or regularities in data, and based on these insights, the system can make predictions, decisions, or classifications. Over time, as more data becomes available and the model adjusts its internal parameters, its accuracy and effectiveness can increase, allowing it to autonomously adapt to new information and evolving situations.

Supervised learning is a method within machine learning where the algorithm is trained on a labeled dataset, meaning that each training example is paired with the correct output. The algorithm’s goal is to learn a mapping from inputs to outputs and make accurate predictions on new, unseen data. The “supervision” consists of the algorithm making predictions and then being corrected by the provided labels whenever it’s wrong. Common tasks in supervised learning include classification, where inputs are categorized into two or more classes, and regression, where the objective is to predict a continuous value.

Unsupervised learning is a type of machine learning where algorithms are trained on data without explicit labels, aiming to identify underlying patterns or structures in the data. Instead of learning a mapping from inputs to outputs, unsupervised learning focuses on extracting insights such as clusters or groupings of similar data points, or reducing the dimensionality of data. Common techniques include clustering, where data is grouped based on inherent similarities, and dimensionality reduction, where data is transformed into a lower-dimensional space while preserving its essential characteristics.

Training data refers to the initial set of data used to help a machine learning model learn and adapt to specific tasks. It consists of input samples and, in supervised learning scenarios, the corresponding output labels. The model uses this data to adjust its internal parameters and make predictions. The quality, quantity, and diversity of the training data are critical, as they directly influence the model’s ability to generalize to new, unseen data, with potential implications for the model’s accuracy and effectiveness in real-world applications.

Synthetic data is artificially generated information, created programmatically, rather than being collected from real-world events. It is often used in contexts where actual data is scarce, sensitive, or difficult to obtain. By simulating features of real datasets, synthetic data can serve as a stand-in for testing, model training, or system validation, ensuring data privacy and facilitating scenarios that might be challenging with real data. When crafted correctly, it allows researchers and practitioners to glean insights, build models, and make decisions without accessing the genuine data sources.

A neural network is a computational model inspired by the way biological neural systems process information. Composed of interconnected nodes or “neurons,” it is designed to recognize patterns and transform input data into an output through a series of layers. Each connection between neurons has a weight, which adjusts during training, allowing the network to learn from input data and make predictions. Neural networks are foundational to deep learning and are employed in diverse tasks like image recognition, language processing, and game playing due to their ability to adapt and refine their internal weights based on the data they encounter.

Natural Language Processing (NLP) is a branch of artificial intelligence that focuses on the interaction between computers and human language. It seeks to enable machines to understand, interpret, and generate human language in a way that is both meaningful and contextually relevant. NLP encompasses a range of tasks including sentiment analysis, machine translation, speech recognition, and text summarization, leveraging various techniques from linguistics and machine learning to bridge the gap between human communication and computer understanding.

Sentiment analysis, often referred to as opinion mining, is a sub-field of Natural Language Processing (NLP) that aims to determine the emotional tone or attitude conveyed within a piece of text. By analyzing words and phrases, it identifies whether the sentiment of the content is positive, negative, neutral, or even more specific emotions like happiness, anger, or sadness. This technique is widely used in business intelligence and social media monitoring to gauge public opinion, customer reviews, or feedback, allowing for informed decision-making and tailored responses.

Deep learning is a subset of machine learning that utilizes neural networks with multiple layers (often called “deep” networks) to analyze various forms of data. These deep networks are capable of automatically discovering and extracting complex patterns from vast amounts of data, making deep learning particularly powerful for tasks like image and speech recognition, natural language processing, and more. The “depth” of the network, characterized by many hidden layers, allows it to learn hierarchies of features, enabling it to handle intricate computations and produce sophisticated predictions or classifications.

A Large Language Model (LLM) is a type of artificial intelligence system designed to understand and generate human-like text based on massive datasets. Trained on vast amounts of text data, these models leverage deep learning architectures, typically transformer-based, to capture intricate patterns and nuances of language. Their size, often characterized by billions or even trillions of parameters, enables them to produce coherent, contextually relevant, and sometimes creative text across diverse topics, making them valuable for tasks ranging from text completion and translation to content generation and question-answering.

Generative AI refers to AI models and systems that can create new, previously unseen content or data based on patterns they’ve learned from existing data. This can encompass generating images, music, text, or any other form of data. Common examples include neural networks like Generative Adversarial Networks (GANs) which can produce realistic images or Deep Learning models that craft coherent text. The essence of generative AI lies in its ability to produce novel outputs rather than simply classifying or analyzing input data.

A Generative Pre-trained Transformer (GPT) is a type of artificial intelligence model primarily used for natural language processing tasks. It utilizes a transformer architecture, which excels in capturing intricate patterns and relationships in data. GPT is “pre-trained” on massive amounts of text, enabling it to generate coherent and contextually relevant sentences. The “generative” aspect denotes its ability to produce new, original content based on its training. With subsequent fine-tuning, GPT can be tailored to specific applications, from text completion to answering questions or generating creative content.

Bias, in a general context, refers to a predisposition or inclination towards a particular viewpoint, often at the expense of alternative perspectives. In the realm of machine learning and artificial intelligence, bias denotes systematic and unfair discrimination in model outputs, often stemming from non-representative training data, flawed algorithms, or subjective human decisions during model design. Such biases can lead to skewed results, perpetuating stereotypes or inaccuracies, and thereby affecting the fairness and trustworthiness of AI systems.

A chatbot is a software application designed to simulate human conversation, either via text or voice interactions. Leveraging pre-defined rules, machine learning, or more advanced natural language processing techniques, chatbots can answer questions, assist users, or perform tasks in a conversational manner. They are commonly deployed on websites, messaging platforms, or other digital channels to provide instant customer support, gather information, or facilitate user experiences without the constant need for human intervention.

A hallucination, in the context of an AI model, refers to the generation or output of inaccurate, fabricated, or nonsensical information that doesn’t align with reality or the data it was trained on. These anomalies arise from the model’s internal representations and biases, often due to overfitting, lack of diverse training data, or inherent limitations in the model’s architecture. Essentially, the AI “imagines” details or connections that aren’t factually correct or contextually relevant, leading to outputs that might be coherent but are not accurate or true.


Understanding the terms associated with AI is the first step to understanding the “nuts and bolts” of AI and understanding how they apply is the next step. For example, ChatGPT is: artificial intelligence that is also a large language model and generative AI. It’s also a chatbot (Chat) and a Generative Pre-trained Transformer (GPT). Now you understand more about what ChatGPT is and what it does!

Now that we’ve defined bias as one of our terms, we will discuss types of bias in AI algorithms and how they can impact the performance and accuracy of the algorithm.

For more regarding Cimplifi specialized expertise regarding AI & machine learning, click here.

In case you missed the previous blogs in this series, you can catch up here:

The “Nuts and Bolts” of Artificial Intelligence for Legal Professionals

The “Nuts and Bolts” of AI: Defining AI

The “Nuts and Bolts” of AI: Types of Bias in AI

The “Nuts and Bolts” of AI: Privacy Considerations

The “Nuts and Bolts” of AI: Transparency, Explainability, and Interpretability, of AI

The “Nuts and Bolts” of AI: ABA Guidance on the Use of AI

The “Nuts and Bolts” of AI: The Current State of AI Regulations

The “Nuts and Bolts” of AI: Current Proven AI Legal Use Cases

The “Nuts and Bolts” of AI: Emerging Use Cases and the Future of AI for Legal

We invite you to stay informed and join the conversation about AI. If you have questions, insights, or thoughts to share, please don’t hesitate to reach out.