Earlier this week I received a question at work about ChatGPT. This morning, my LinkedIn feed is filled with buzz about ChatGPT. So what exactly is ChatGPT and what is all the buzz about?
Chat GPT is a chatbot system by OpenAI. It is a language model that interacts with users in a conversational manner to respond to inquiries and requests. The model is currently in a research phase, available to the public free of charge. OpenAI's website provides context on how the model was trained using reinforcement learning. Reading that, I was having flashbacks to undergrad psych classes learning about Pavlov's dogs. OpenAI also provides visibility into a number of the model's limitations, including:
- Providing inaccurate answers
- Failing to ask clarifying questions to understand the user's intent
- Exhibiting biased behavior
Mike Jacka wrote a blog post for the Institute of Internal Auditors about a potential use case for ChatGPT to write audit reports, which could really reduce our time spent drafting audit reports in the future--saving organizations loads of time!
Curious, I started playing with ChatGPT's functionality. I asked it a number of questions to see how it would respond. The model generated responses immediately after I submitted the questions. It felt kind of like conducting a digital interview, since I was asking the model questions about itself.
I will share portions of the "interview" I had with the tool below, but first I want to put a disclaimer out there. I have not independently verified the accuracy of the model's responses. I also did not perform due diligence to gain assurance over controls to ensure the accuracy of the responses provided. The point of this post is to show how far this technology has come and to highlight a potential use case. The purpose of the post is not to provide guidance on the proper use of this technology.
Now that we have that out of the way, here are the questions I asked and the model's responses:
What are use cases of ChatGPT in the financial services and insurance industries?
There are several potential use cases for chat GPT (generic pre-trained) systems in the financial services and insurance industries.
One possible use case is in customer service. Chat GPT systems could be used to provide instant assistance to customers, answering common questions and helping them navigate the company's products and services. This could potentially reduce the workload of human customer service representatives, freeing them up to handle more complex or specialized inquiries.
Another potential use case for chat GPT in the financial services industry is in fraud detection. Chat GPT systems could be trained to analyze customer communications and identify patterns or anomalies that may indicate fraudulent activity. This could help financial institutions detect and prevent fraud in real-time, improving their overall risk management capabilities.
In the insurance industry, chat GPT systems could be used to assist with the underwriting process. By analyzing customer data and information about potential risks, chat GPT systems could help insurers determine the appropriate coverage and premiums for individual customers. This could help insurers streamline the underwriting process and make it more efficient.
Overall, chat GPT systems have the potential to improve efficiency and effectiveness in a variety of applications within the financial services and insurance industries. By leveraging the capabilities of these systems, organizations in these sectors can better serve their customers and protect against potential risks.
How might ChatGPT be used to increase efficiency and effectiveness in the internal audit process?
There are several ways in which ChatGPT (generic pre-trained) systems could potentially be used to increase efficiency and effectiveness in the internal audit process. Here are a few examples:
Automating routine tasks: ChatGPT systems could be used to automate routine tasks in the internal audit process, such as reviewing and analyzing financial statements or other data. This could potentially free up auditors to focus on more value-added activities, such as developing audit plans or conducting more in-depth analysis.
Providing real-time assistance: ChatGPT systems could be used to provide real-time assistance to auditors as they perform their duties. For example, a ChatGPT system could be trained to answer common questions or provide guidance on complex issues, allowing auditors to work more efficiently and effectively.
Enhancing communication and collaboration: ChatGPT systems could be used to facilitate communication and collaboration among auditors, as well as between auditors and other stakeholders. For example, a ChatGPT system could be used to help auditors share information, collaborate on audit plans, or coordinate with other teams or departments.
Overall, ChatGPT systems have the potential to significantly improve the efficiency and effectiveness of the internal audit process by automating routine tasks, providing real-time assistance, and enhancing communication and collaboration. By leveraging the capabilities of these systems, organizations can ensure that their internal audit processes are efficient, effective, and add value to the organization.
What risks should organizations consider when using ChatGPT?
There are several risks that organizations should consider when using chat GPT (generic pre-trained) systems. These risks may vary depending on the specific use case and context in which the chat GPT system is being deployed.
One risk to consider is the potential for errors or misstatements in the output of the chat GPT system. This could occur if the system is not properly trained or if it is fed incorrect or incomplete data. It is important for organizations to implement controls to ensure the accuracy and reliability of the output of the chat GPT system, such as regularly testing and validating the system and reviewing its output for errors or discrepancies.
Another risk to consider is the potential for unintended consequences or ethical issues. For example, if a chat GPT system is used to generate content or responses, it is important to ensure that the output is appropriate and does not contain offensive or inappropriate material. It is also important to consider the potential for bias in the training data used to develop the chat GPT system, as this could result in biased or discriminatory output.
A third risk to consider is the potential for security breaches or unauthorized access to the chat GPT system or the data it uses. Organizations should implement appropriate security controls to protect against these risks, such as secure data storage and access controls.
Overall, it is important for organizations to carefully evaluate and manage the risks associated with using chat GPT systems to ensure the integrity and reliability of their operations.
When using ChatGPT, what is the user accountable for and what is OpenAI accountable for?
When using ChatGPT (generic pre-trained), the user of the system is generally responsible for determining how the system will be used and for implementing appropriate controls and safeguards to ensure the integrity and reliability of the system. This may involve setting up and training the system, as well as regularly reviewing and testing its output to ensure it is functioning as intended.
OpenAI, as the creator and provider of the ChatGPT system, is generally responsible for ensuring that the system is of high quality and operates as intended. This may involve providing support and documentation for the system, as well as regularly updating and improving the system to ensure it remains effective.
Overall, it is important for users of ChatGPT to understand their responsibilities when using the system and to work closely with OpenAI to ensure that the system is used effectively and appropriately.
Finally, I asked the model how to properly cite the output of ChatGPT by OpenAI. Within seconds, it produced a template from which I created the following citation: OpenAI Chat GPT. Retrieved 1/7/2023, from https://chat.openai.com/
As I mentioned earlier in this post, I haven't independently verified the accuracy of these responses, nor have I performed due diligence to gain comfort over the accuracy of the responses provided. The point of this post was not to provide guidance on the proper use of this technology, rather it is to show an example of how far the technology has come and how easy it is to use. I suspect we will be seeing a lot more of this technology used pervasively across many industries.
I'd love to hear what you have to say on the topic. How have you used this technology and what have you learned?