What is AI?
Artificial intelligence is the science of making machines that can learn, similar to the ways humans do. AI technology can process large amounts of data in ways that humans can’t. Machines and programs that incorporate AI in their design are able to recognize patterns and make decisions based on their experience of a dataset and therefore their ability to predict what would come next in a pattern.
How does AI work?
From the SAS institute: “AI adapts through progressive learning algorithms to let the data do the programming. AI finds structure and regularities in data so that algorithms can acquire skills. Just as an algorithm can teach itself to play chess, it can teach itself what product to recommend next online.”
What is AI good for?
Artificial intelligence has been a subject of study since the 1950’s. Practical applications for this method of programming have varied over time. Finally, we live in a time with enough data to make use of AI’s ability to recognize and predict patterns. So far many of the applications for using artificial intelligence in healthcare are behind the scenes and not available to the general public. The idea is that by training these models on data belonging to patients with health conditions, a program might be able to recognize correlation between patterns in data sets that humans haven’t noticed and therefore lead to earlier diagnostic capabilities and all sorts of insights. Most people are not familiar with the uses for AI in healthcare unless they are studying that science specifically. But there is one common application of AI that most people are familiar with, and that is the language model.
What is a language model?
Programs that make use of an AI language model are designed to generate language that makes sense to a human reader. By having a “conversation” with the language model, the program gathers information about the language it is given and it returns language that should make sense given the context of the topic provided and all the other conversations and articles (language data) it has been trained on.
The fascinating effect of successfully creating an AI language model, especially a chat bot, is that the user experiences the sensation that they are speaking with an intelligent being. This effect is similar to face pareidolia – the phenomenon of seeing faces in everyday objects. Depending on the topic the chatbot is generating language about, the user could experience a very emotional and seemingly intimate moment with a machine because of the generated language.
Do language models generate truth?
As far as the author of this article understands it, there is no correlation between the content of the language generated by a chatbot and the accuracy of the information being communicated. The goal of the program is to make sense grammatically and contextually; to use words that the reader will relate to and potentially find useful. It’s important to remember that even though artificial intelligence is generating language, that doesn’t mean that it is thinking about that language as a representation of our world. It’s only facet of intelligence is based on creating more language that fits a variety of feasible patterns of communication. There was a time in the language model’s existence when nothing it generated made sense, and over the course of its learning experience it was able to learn how to pick up on increasingly complex uses of language. Whether or not the words generated are true should always be verified and not taken at face value. Some truths are logical or emotional, but when it comes to information, the scientific method requires rigorous review before it makes facts of language that makes sense in theory.
What is an AI chat bot good for?
Well-trained language models can be used for anything that requires communication, but the accuracy of the information shouldn’t be left up to the artificial intelligence to determine. A good use of this tool would be to summarize existing articles or to write a first draft of a document with all the basic information provided by a human. Users can tell a language model exactly what they are looking for as far as tone, and they can ask the AI to rewrite sentences and create summaries which is useful for creating HIPAA Policies and Procedures for Privacy and Security of PHI. Just keep in mind that the goal of the program is to make grammatical and emotional sense- it doesn’t measure factualness.
What is an AI chat bot not good for?
Artificial intelligence means machines can learn, it doesn’t mean machines can know. Just because something is written down and makes sense doesn’t make it true. Using a chat bot to create policies and procedures for HIPAA privacy and security from start to finish would be very similar to adopting policies from another company or purchasing a non-customized template. It’s really just lip service to HIPAA, where as filling out a risk assessment and creating your policies from that is actual compliance.
Read more about adopting policies in this article: How to Fail a HIPAA Audit
Upcoming Webinar: Join us September 20th at 2pm EST to discuss how AI is being used in HIPAA
For step by step support and guidance in HIPAA compliance, consider trying HIPAA Prime. You’ll gain access to our team of experts who will create policies and procedures specifically for your business. Book a Clarity Call today to learn more about HIPAA Prime!
Further Reading on AI:
- https://hai.stanford.edu/news/language-models-are-changing-ai-we-need-understand-them
- https://medium.com/@diamond_io/artificial-intelligence-101-everything-you-need-to-know-to-understand-ai-8e20fe4bd750
- https://www.simplilearn.com/tutorials/artificial-intelligence-tutorial/what-is-artificial-intelligence#:~:text=Artificial%20Intelligence%20is%20a%20method,develops%20intelligent%20software%20and%20systems.
- https://www.linkedin.com/pulse/artificial-intelligence-101-understanding-basics-ai/
