A seemingly overnight boom of artificial intelligence (AI) is sending the world into a spin. While pundits discuss the benefits and risks of AI, businesses are investigating how and where to add automation into their workflows — including communications.
AI chatbots, such as ChatGPT and other natural language processing (NLP) programs, have created new opportunities, challenges, and cautions. The health care industry in particular must guard against errant content, always vetting computer-generated communications with a human eye for accuracy.
ChatGPT stands for “chatbot” using the “generative pretrained transformer” large language model. The Transformer was created by Google and released in 2017.
The original goal of the Transformer was to make it easier to predict the next word in groups of text. This kind of technology is behind the predictive text word suggestions that appear when we type on our phones, laptops, or other devices. The algorithm understands context and guesses what we are most likely to say next.
Fast forward and we now have powerful AI chatbots from ChatGPT (OpenAI), Bard (Google), Bing (Microsoft), and numerous startups. All of these bots use large language models (LLMs), defined by computer chip maker NVIDIA as models that “recognize, summarize, translate, predict and generate text and other content.”
LLMs are what turn written chatbot prompts into code and match the code to words on the internet to fulfill the prompt.
AI chatbots have valuable use cases in businesses of all kinds. Many companies herald the technology as a thought starter and solution for repeatable and high-volume needs, such as:
Still, the technology requires people — to write prompts, establish content flow, edit for tone and style, tweak for brand voice and vision, and run through quality assurance.
There’s a lot to consider about chatbots and business communication needs. Knowing what they can’t do is just as important as knowing how they can help. AI chatbots:
The algorithms behind the LLMs that power AI chatbots can only pull from recorded information, data, and conversations. This includes social media and other unregulated stores of content.
Every prompt entered into a chatbot — and every answer delivered — becomes part of the larger record. The algorithm learns from each interaction, training itself on user requests without tracking where the data goes or how it gets used.
AI has incredible potential in science and medicine, with the ability to cull through endless amounts of data to find new approaches to disease.
However, when it comes to talking about health care, AI can be harmful. The algorithms are not yet sophisticated enough to determine factual information from untrue statements.
The AHIMA Foundation, a charitable affiliate of the American Health Information Management Association, provides resources for navigating misinformation. Responsible health care companies will vet every word for scientific validity, ensuring evidence-based content that providers and patients can trust.
“Since there was no legal review request form and because it appears from the disclosure that this is for the health plans, then the Highmark leader should be from the health plan and the quote should speak about the topic from a health plan perspective. “ - Christine McMillan
Chatbots and other AI tools are teaching themselves through machine learning (ML) as people interact with the technology.
Discussions on ethics and efficiency will continue to swirl, while businesses learn to balance their needs with their customers’ preferences and demands. Every use case will be different, as will the guardrails.
AI chatbots have cemented their place in our modern world. Our responsibility is to use them wisely to make tasks easier — without losing our human touch.
All references to “Highmark” in this communication are references to Highmark Inc., an independent licensee of the Blue Cross Blue Shield Association, and/or to one or more of its affiliated Blue companies.
Talk to your client manager today and we’ll help you find the right health plan for your company.
