No technology in recent memory has generated more hype, both positive and negative, than ChatGPT. ChatGPT is probably the most well-known of many generative artificial intelligence (AI) applications and tools that are currently available. There are many tasks that it can help to automate, optimize, and make more efficient. But it also has limitations and risks and cannot entirely replace most human work.

To get the most out of ChatGPT, business owners should identify areas in which AI can supplement human work, but steps should also be taken to avoid risks such as inaccuracy, confidentiality, and bias.

What Is ChatGPT?

According to its creator, the San Francisco start-up OpenAI, ChatGPT is an AI-powered language model that generates original, human-like text based on context and past conversations.^[1] The latest version, GPT-4, was released on March 14, 2023. GPT-4.5 is expected to be introduced in October 2023.

The “GPT” in ChatGPT refers to the learning model it uses—Generative Pre-trained Transformer. It is pretrained on a huge amount of publicly available data from the internet. When a user inputs a message, ChatGPT feeds the text into a language generation component that is used to create a response. ChatGPT is also able to learn and adapt to user preferences to provide a more personalized experience with continued use.

How Small Businesses Can Use ChatGPT

Since OpenAI released a ChatGPT demo in late 2022, it has been used for everything from writing essays, computer code, and fiction to travel planning and customer support.

Although there have been dire predictions that AI technology such as ChatGPT could replace hundreds of millions of jobs by automating midlevel work, some are pumping the brakes about what it is and what it can do. Even OpenAI CEO Sam Altman has cautioned that ChatGPT is “incredibly limited.”^[2]

Despite its limitations, ChatGPT can help perform a variety of tasks, including the following:

• Customer service. ChatGPT can automate responses to customers who ask common questions, analyze customer feedback to identify areas that should be prioritized, and produce customer emails.
• Generating summaries and outlines. Need to summarize meeting notes or come up with an outline for a presentation or a project? Having difficulty understanding a complex concept? ChatGPT can generate summaries and outlines in seconds.
• Translation. Translating from one language to another with ChatGPT is simple: just provide the text you would like translated and specify the language you would like the text translated into.
• Coding and SEO. ChatGPT can write basic computer code and generate search engine optimization (SEO)-friendly keywords for a topic.
• Tech and software support. ChatGPT can perform basic IT troubleshooting and save money on technical issue support. It can also provide natural language answers about how to use apps, programs, and technology.
• Content creation. Use ChatGPT to find article topics, create outlines, brainstorm, and draft marketing and sales copy.

ChatGPT Risks

Although ChatGPT and other AI technologies offer many potential benefits, they create inherent risks as well.

ChatGPT’s limits stem from the way it works. Because it relies on information from the internet and human feedback, outputs can be inaccurate, incomplete, or biased.

There is a saying in the computer world: garbage in, garbage out. Language models absorb as much information as they can from the web and then use it to generate new text. Sometimes their answers sound convincing but are wrong—what AI researchers call hallucinations.^[3]

ChatGPT may lack detailed or technical knowledge about a topic. This might not be as risky if it is used merely to create an outline. However, the inclusion of incorrect information in a client-facing document could be embarrassing, or worse, lead to the loss of business. Accurate information that is inadvertently disclosed, though, is another significant concern. If employees enter confidential company information into ChatGPT, they could accidentally  disclose trade secrets and other proprietary data. On the flipside, the language model could generate results that include protected intellectual property, which could lead to inadvertent copyright infringement.

And while ChatGPT could be used to remove unconscious human bias from certain decisions, bias can still exist as a result of the information it was trained on. Used in connection with employment decisions, ChatGPT could therefore inadvertently discriminate on the basis of protected characteristics (e.g., race, color, religion, sex, or disability), creating a risk of legal liability.

The risks of ChatGPT should not stop small business owners from embracing AI automation to save time and money. But it is far from a one-size-fits-all solution. Used in the wrong context or without proper safeguards, ChatGPT could create significant problems for your business.

Companies that use ChatGPT should create and implement policies that clearly address technology use, codes of conduct, and confidentiality and trade secret agreements in the context of using AI tools. Given the rapid pace of change in the AI field, these policies should be updated frequently, and employers should consider training employees on the latest uses, benefits, and risks of ChatGPT.

For assistance in developing an AI policy for your small business, contact King & King at 251-501-6100 or 731-664-6864 to set up a consultation.

^[1] ChatGPT Overview, OpenAI, https://chat.openai.com/ (last visited Oct. 13, 2023).

^[2] @sama, X (Dec. 10, 2022, 6:11 PM), https://twitter.com/sama/status/1601731295792414720?lang=en.

^[3] Cade Metz, What Makes A.I. Chatbots Go Wrong?, N.Y. Times (Apr. 4, 2023), https://www.nytimes.com/2023/03/29/technology/ai-chatbots-hallucinations.html.