What is AI and how does it work?

AI, which stands for artificial intelligence, is a broad term encompassing a wide range of technologies. The AIs most talked-about these days are called generative AIs or LLMs (Large Language Models).

Generative AIs use a technology which allows them to generate many different kinds of content, including text, images, code, and others.

What's the technology behind generative AI?

A generative AI draws from a vast sets of text and images—known as a data set—to generate new content.

It does this with a large network of computer-hosted components known collectively as an artificial neural network (ANN). Artificial neural networks are modeled loosely after the human brain which can store, act on, and transmit billions of small pieces of data in myriad, intricate patterns.

The technology at the core of a generative AI is a Large Language Models (LLM). The word "model" refers the types of algorithm (i.e. set of calculations) which allows it to process data in a particular way.  

Here the word "model" is being used the same way it's used to describe computer systems that can model weather systems, biological systems, or any complex mathematically-based system.

In the case of generative AIs what's being modeled is the comprehension and creation of language.

drawing of a human brain scan next to an illustration of an artificial neural network

Although a generative AI does not truly replicate the biological processes in a human brain, it can emulate (i.e. model) a specific cognitive skill called "pattern recognition".

A generative AI develops the ability to detect and create meaningful patterns out of data when it is given a very large set of data to reference and is provided with a very large amount of computational power.

With enough data and enough "compute", generative AIs develop the ability to identify and generate meaningful language, images, code, and other types of content.

What is generative AI good at?

Generative AI is good at recognizing and replicating patterns in the various languages contained within its data set.

Some AIs, such as ChatGPT, have been trained on an enormous data set which contains many different kinds of languages, including, for example human languages like English, Spanish, and Hindi, computer languages like Python and C++, and visual languages, like x-rays and sports photos.

Because ChatGPT has been trained with data that includes all these languages, it develops the ability to generate content in all of those languages too.

Why are some AIs like ChatGPT remarkably good at generating text?

"'GPT' stands for 'Generative Pre-Trained Transformer'. A transformer model is an innovation on a more conventional neural network that adds in 'attention layers', allowing the model to selectively 'pay attention' to different positions of an input string. So, for example, if the model is 'learning' from an example sentence, different layers of the neural network can 'attend' to different words in the sentence, and can learn about the relationships between different parts of the sentence, not just those parts that are immediately adjacent. Models like GPT-3 have an enormous number of 'parameters', which allow the model to 'attend' to many, many aspects of the 'context' in which a particular word appears. This architecture gives transformer models a way of flexibly representing a LOT of information about the context in which different words appear, and that flexibility is a big part of what gives transformer models the power they have to generate such fluent-sounding text."

Source: Against the use of GPTZero and other LLM-detection tools on student writing by Whitney Gregg-Harrison

A word of caution about generative AI security —

No private data should be input into a generative AI.

AIs are new and experimental technology. Even AIs which claim to be secure may have security vulnerabilities which have not yet been discovered or addressed.

For this reason we strongly recommend that you only share content with an AI that you would be willing to share publicly online. Never share confidential data about individuals or organizations, including names, email addresses, or other identifying or sensitive information.

Before you use AI for teaching see: Tufts Guidelines for Use of Generative AI Tools