2.1 AI

To define “AI”, it is important to note that:

“Al is a broad umbrella term with no single meaning […] The Organization for Economic Cooperation and Development (OECD) defines an “Al system” as a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments.”¹

Generally speaking, AI can be understood as computer systems able to independently complete tasks that would otherwise require human intelligence and intervention.

Due to the lack of consensus regarding definitions, “the meaning of the term Al is contextual and may be defined differently in legal instruments, policy settings, or in contracts as part of a description of goods or services. Thus, legal requirements, contractual promises, and dialogue that refer to Al should be understood and interpreted with reference to how the term is used in the specific context.”²

Perhaps a more helpful working definition is the following:

Artificial intelligence (AI): The capacity of computers or other machines to exhibit or simulate intelligent behaviour; the field of study concerned with this.³

Certain forms of AI can automate repetitive tasks and help lawyers streamline their legal processes.

Newer forms of AI can produce content in response to plain language prompts and make recommendations, making them increasingly popular tech tools. The latest form of AI that has created enormous interest is generative AI, described below.

2.2 Generative AI

One helpful definition of generative AI is the following:

“Generative AI is a subcategory of AI that uses deep learning algorithms to generate new outputs/ content based on large quantities of existing or synthetic (artificially created) input data.”⁴

Generative AI differs from the types of AI that lawyers are accustomed to using due to its ability to create new content based on its training data. In other words, when using generative AI, the technology not only analyzes or categorizes existing information; it uses patterns and data it has learned from existing content to generate new content that is similar in style and structure to the original data but entirely new.

Like all forms of AI, it cannot understand its output and meaning in the same way that humans do.

This means that generative AI cannot independently validate or audit the accuracy of its results. It may confidently and misleadingly supply false outputs (commonly known as “hallucinations”), as noted repeatedly in guidance documents and communications from regulators. ⁵  Essentially, when it is ‘hallucinating’, generative AI makes up information or content that does not align with reality or factual accuracy.

Lawyers and firms should also understand that generative AI tools not only process data to generate responses – i.e. create new content - but also use the data to enhance or further “train” the system itself. AI tools can only use data that exists and is only as good as it’s data source.

It is also important for lawyers and firms to understand that generative AI was not originally designed for use in legal practice or designed to undertake specific legal tasks with appropriate responses built in to mitigate risks inherent with its use.

2.3 Large Language Models

A large language model (LLM) is an AI system trained on an especially large amount of data. Large language models are one form of generative AI.⁶  The primary task of LLMs is to predict the next word in a string of words, with the result that LLMs can generate convincing but inaccurate information.

2.4 Chat GPT

Open AI’s Chat GTP is a popular AI LLM that can respond to natural language prompts and confidently generate responses that generally appear appropriate.  It is a conversational generative AI system fine-tuned for dialogue tasks.  Various versions have been developed.  Earlier versions were available without cost and were trained on data that had been “scraped” from the internet up to a fixed period. A newer version is available to subscribers and information may be accessed in real-time.  Notably, Chat GPT is only one example of an AI LLM.  Others are available and continue to be refined.