Authored by: Mark Adams, IOM,SPHR,SHRM-SCP, Director of Compliance
ChatGPT has captured the imagination of many. For some, it is the tool that they have been waiting for. The catalyst to generate content when one encounters “writer’s block”. A technological marvel which upon providing a short query and the click of the button can appear to solve any problem. Is it all that it is cracked up to be? Is it too good to be true? Is it a friend for HR or a foe?
Well, it is not an understatement that ChatGPT is evolving rapidly, but contrary to popular belief, it didn’t evolve overnight. It has been several years in the making. While it went mainstream in 2022, its origins go back to 2018 when “ChatGPT-1” was launched. Over time, ChatGPT’s evolution has been in large measure based on the infusion of more information or “parameters” into the application for it to “learn”. In 2018, that started with approximately 117 million parameters. In 2022, that spiked upwards to 175 billion. With the advent of ChatGPT-4 rolling out this year, it is over 1 TRILLION parameters!
More information. More settings that dictate how it generates text. As it evolves, its responses will be more precise and effective.
Yet what are the limits? What does HR have to be mindful of if it embraces ChatGPT? There are a few considerations.
First, HOW will it be used? There is the version of ChatGPT that is out there for anyone to tap[MW1] into. Simply by setting up a free account which allows an individual to generate content from the information (parameters) that has been fed into the system by its creator – OpenAI.
However, companies can also partner and connect their own information into ChatGPT[MW2] . Here, ChatGPT can “learn” with a company’s own information to meet a company’s specific needs. Some companies, like Coca Cola, Salesforce and others are jumping on the bandwagon with some using it to meet their customer service and performance management needs (after all at its very core it is a “bot”).
Many have postulated that in the world of HR, its immediate benefits will be in assisting or even replacing more “tactical” tasks or functions. They may include the likes of:
- Generating job descriptions
- Generating interviewing questions
- Assisting with screening of candidates
- Developing policies and forms
- Q and A for Managers and Employees – “HR Chatbots”
- Training – create training plans and materials
But checks and balances should be put into place. Despite its algorithms – there have been reported hiccups some of which I detail below
Also just like a house is only as good as its foundation, its utility is only as good as the parameters that are fed into its application. Thus, if the foundational information or parameters that are fed into Chat GPT are not accurate or are biased in the first place, it can have the potential of providing summaries or narrative that are inaccurate or biased as well.
It is out of this concern that there has been a joint statement published by the Federal Trade Commission (FTC) and Equal Employment Opportunity Commission (EEOC), warning employers that they are responsible, and can be liable, for their actions if the use of this technologies create illegal or discriminatory outcomes.
Then there is the question of intellectual property – who owns the output? Is it the end user? If it is based off of knowledge that itself is copyright protected – do the copyright owners have some ownership over the applicable outcome or application?
Section 3(a) of the end user agreement with OpenAI points to the user having ownership but we have yet to see courts weigh in on this matter (after all – it also says that this is the case “to the extent permitted by applicable law”).
Finally, there is the question of confidentiality. Bear in mind that companies who link ChatGPT to its own data or information provide permission for that information to be used by OpenAI and ChatGPT in the future “to improve services” (see Section 3(c) of the end user agreement). This turned out to be an issue for Samsung in a widely covered news story in which a Samsung engineer had uploaded sensitive confidential and proprietary code onto the platform.
Other companies have taken positions on the subject with JPMorgan reportedly restricting use of ChatGPT among staff and Amazon this year reportedly warned employees not to upload confidential information, including code, to ChatGPT.
So, for HR, it is important to be sure management is all on the same page when it comes to tapping this resource and what type of company information may be used (or not used) in this regard.
Then, of course, even if you have established protocols around how to access and use it, there still is the question of vetting out its content when used to ensure it is effectively meeting your business needs from a strategic and cultural perspective. For instance, as any HR professional knows when it comes to policies, not all are alike. There are differing approaches that can be taken on a given subject. Just because ChatGPT generates one for you – it does not mean that it is something that you can automatically “plug and play” into your operations.
If you query Chat GPT “why do employers want to use ChatGPT?”, it will give you various reasons: customer support, scalability, cost efficiency, 24/7 availability, speed. Effectiveness and “fit” are not among the typical responses. This is where HR needs to step up to the plate in today’s technological world. To shine and be that STRATEGIC partner.
So, is it a friend or foe? The answer is that ChatGPT is a tool. It is not a replacement for HR. It is helping your organization understand HOW to use the tool that in the end will demonstrate your value to your organization and separate one HR professional from the other.