What is AI?

Machine Learning

Machine Learning (ML) refers to the field and practice of using algorithms that are able to “learn” by extracting patterns from a large body of data. Types of learning include:

• Supervised learning uses labeled datasets to train algorithms to classify data or predict outcomes. An example of supervised learning is speech recognition.
• Semi-supervised is a machine learning technique that uses a small portion of labeled data and lots of unlabeled data to train a predictive model. An example of semi-supervised learning is a large language model, like ChatGPT.
• Unsupervised learning uses unlabeled data. From that data, it discovers patterns that help solve clustering or association problems. An example of an unsupervised model is a recommendation system, like Netflix.
• Reinforcement learning sequence of successful outcomes will be reinforced to develop the best recommendation for a given problem. Examples of AI systems relying on reinforcement learning are robots and video games.

AI solutions use one, or in some cases several, of these ML techniques.

Generative AI

Generative AI is a type of algorithm capable of generating text, images, or other media in response to prompts. One example of this type of algorithm is chat-based generative pre-trained transformer, like ChatGPT. A pre-trained transformer ChatGPT is a type of AI system that has been pre-trained or pre-loaded with data and that can transform a given prompt into a response by predicting and responding with the group of words most likely associated with the prompt. Generative AI, like ChatGPT is a system built with a neural network transformer type of AI model that works well in natural language processing tasks. In this case, the model:

• can generate responses to questions (generative); 
• was trained in advance on a large amount of the written material available on the web (pre-trained); and
• can process sentences differently than other types of models (transformer).

Training an AI Model

AI model training refers to the process of inputting data into an algorithm, running the algorithm, examining the results, and adjusting the model output to increase accuracy and efficacy of the predictions made by the algorithm. To train for the best outcomes, algorithms need massive amounts of data that capture the full range of what’s possible within incoming data.

With enough training, the algorithm or set of algorithms within the AI model will represent a mathematical predictor for a given situation that builds in tolerances for the unexpected while maximizing predictability.

AI Error

AI models are programmed with human data. Data input into any AI model should be collected from a representative sample of the population. Not all data about all humans in the population is entered into the model. Equally, humans are not 100% predictable. AI models contain errors and can be imprecise. The margin of error in AI models is related to the distance of the model from the data points included in the model. The lower the amount of error in the model, the higher the predictive power of the model. However, no model is 100% correct. Thus, AI can produce “hallucinations” or incorrect predictions.

Examples of AI

ChatBots use generative AI and natural language processing to simulate human-like conversations in a chat window where the user can ask the bot to help with a variety of tasks, including editing or writing emails, essays, code, and more.

Natural Language Processing (NLP) is the field of AI where computer science meets linguistics to allow computers to understand and process human language.

A recommendation system (or recommender system) is a class of machine learning that uses data to help predict, narrow down, and find what people are looking for among an exponentially growing number of options.