ChatGPT & Generative AI: Everything You Need to Know

What is 'Generative AI'?

ChatGPT is one of a suite of AI-powered technologies that is being dubbed "generative AI." These are tools that can take a prompt or query from a user (the "input") and respond to it with a type of "output" that resembles what a human would create. These tools are referred to as "generative" because they do not rely on a database of preformulated answers or responses that they can retrieve to address user input. Instead, they have been trained to "recognize" a user's input and to "generate" a response entirely on their own.

Some of the more well-known examples of generative AI include:

• ChatGPT. ChatGPT is — in simplified terms — a powerful chatbot. It is a "large language model" powered by a neural network that can: a) receive natural-language input from a user; and b) provide natural-language output that resembles how a human would respond. ChatGPT is operated by the company OpenAI.
• Generative AI for Images. There are also generative AI tools that autonomously create images. OpenAI operates DALL-E a tool that creates images from a natural language description. The tool "Stable Diffusion" is similar, creating images automatically in response to natural-language inputs from users.
• Generative AI for Coding. Generative AI tools also assist with creating computer code. OpenAI's Codex uses an AI model to generate computer code in response to user input. Codex also powers Github's "Copilot" functionality, which suggests code to programmers in real time in response to the code they are already creating.

Common Generative AI Use Cases

Generative AI tools are not restricted to any particular use case. But requests to corporate legal departments seem to be presently coalescing around several specific use cases:

• Coding Assistant. ChatGPT is reportedly a serviceable coder.  It may not be able to create "ready-to-deploy" code, but it can take a goal ("Write a C++ script that does X") and generate a workable first draft. Generative AI tools like ChatGPT, Codex, or Copilot can potentially save hours of coding time per job — and thus be valuable to architects, developers, and others who are tasked with shipping products and features on deadlines.
• Content Creation. Generative AI can quickly create a wide variety of content. For example, ChatGPT can write draft copy for Sales, Marketing, or Comms. Similarly, image generators like DALL-E or Stable Diffusion can quickly generate a series of images that could be used to mock up initial versions or features of mobile apps, games, or other visual-heavy products or services.
• Document Drafting. ChatGPT can draft documents upon request. For example, it could draft policies for HR. It also drafts legal documents if requested (even though these come caveated, stating that a lawyer should be consulted). Again, ChatGPT's output is not necessarily ready-to-use, and its quality has not yet been comprehensively reviewed. But, it could conceivably be viewed as a sort of virtual "draughtsman" able to generate initial drafts for review.
• Customer Support. AI-powered chatbots are already a feature of customer service. Some AI-powered chatbots still rely on a limited set of preformulated answers. ChatGPT does not; it generates answers that do not exist prior to a query, based specifically on what it is asked. Customer support departments are thus evaluating whether ChatGPT can be used to improve operations. Some companies may be considering how to integrate ChatGPT directly into customer-facing interactions. Still, even if ChatGPT isn't used to generate responses directly to customers, some companies may be considering how it can streamline support operations. For example, ChatGPT could quickly review the history of a customer's prior interactions with a company — then summarize this in 3-4 bullet points for a customer service rep.

Legal Considerations When Using Generative AI

While business clients may be seeing benefits from using generative AI, corporate counsel tends to focus on legal risks that may arise from permitting enterprise use. The risks of generative AI are still being discovered, so this advisory cannot present an exhaustive, closed-ended list of considerations that may be relevant to counsel. At present, however, reporting has identified several relevant considerations. Some of the more salient are: