Part of growing a business means reducing how much work we have to do on tasks unrelated to our expertise. As therapists, our expertise is therapy, and so it is often a good idea to utilize tools – and people – to try to reduce how much time and energy we spend on those tasks so that we can focus on delivering better client care and putting our minds and resources on what we’re best at.
You’ve probably heard about AI – about tools like ChatGPT – and how they can help handle a lot of your work for you. Some of it is true. Some of it is not. Someday, I’ll create a very long, detailed post or series of posts that goes into AI in greater detail. But for now, I want to take a moment to explain what “AI” is in the context of tools like ChatGPT, and whether it’s something worth exploring for your business.
The Basics
For the purposes of this post, we’re going to be referring to ChatGPT, though all AI tools work essentially the same. We’re also not going to debate the ethical or climate impact of using these tools. We can save those debates for another time.
ChatGPT is what’s known as an LLM, or “Large Language Model.” It is, at its core, and algorithm. OpenAI, the company that created ChatGPT, has created a database of trillions upon trillions of text-based files, and when you ask the AI a question (known as a “prompt”) or tell the AI tools a task you need completed, the system’s algorithm determines the answer based on its understanding of the database.
The result is that you can tell ChatGPT something like “How should I begin an email to my boss letting her know I’m quitting to open a private practice” and it will generate an answer using its database, such as:
“I hope this message finds you well. After careful consideration, I have decided to take the next step in my professional journey by opening my own private practice. This decision has not been easy, as I have greatly appreciated the opportunity to work under your leadership and alongside such a dedicated team.”
If you want, you can create something more fun. You can write something truly oddball, such as “write me a poem about cilantro in the style of Robert Frost” and it will generate a response:
Among the Leaves of Cilantro Green
By frost’s first breath, the leaves remain,
Cilantro green, a subtle vein.
Not of the bold, nor faintly shy,
But speaks to those who pass it by.
The taste, divisive, sharp, and true,
To some, a joy, to others, rue.
The earth it holds within its leaf,
A fleeting burst, then swift relief.
In gardens wide and kitchens small,
It stands beside the rest and all.
A hint of warmth, a trace of cool,
Cilantro’s charm is calm, yet full.
The hand that picks it feels its weight,
A simple herb, yet draws debate.
For some, a fragrance fresh and bright,
For others, soap upon first bite.
And so, like frost that bends the trees,
It sways and bends in every breeze.
But leaves its mark, both soft and clear,
Cilantro whispers, “I am here.”
As you can see, its ability to generate text is incredible, and it takes quite literally seconds. You can have a conversation with it. You can ask it to write a eulogy. You can have it edit writing that you’ve already created. It will use its algorithm to generate something that is typically grammatically perfect or near perfect, and sounds exactly like something you can imagine a human has written.
That means that you can – at least theoretically – use ChatGPT to compose email responses, summarize homework assignments, and even ask it what documents you need to file with the state in order to set up your LLC. All of those make it a potentially useful tool for efficiency and possibly even cost savings, depending on what you use it for.
AI Is NOT Artificial Intelligence
It should be noted, however, that while we call it “AI,” it is NOT artificial intelligence. Intelligence requires thought, and ChatGPT is incapable of thinking. What this modern version of AI is, is a very advanced generative text algorithm.
If you have a smartphone and you’re composing a text message, you’ve probably noticed that right above the text section is a guess about what your next word might be. If you click on it, it will suggest another word, and then another, and eventually you’ll have something that appears sort of sentence-like but is grammatically pretty terrible and mostly nonsense. This button has been on Smartphones for 10+ years and its specifically known as the “Predictive Text” button.
ChatGPT is sort of like of like that, but a million times more advanced. It doesn’t know what it’s saying, because it doesn’t have the concept of thought nor an understanding of words, but it does have a highly advanced algorithm that is capable of determining what the next word should be based on the words before it in order to create a piece of content that reads like a human wrote it.
Note: You’ll sometimes see articles online that say things like “we asked ChatGPT which of these three stocks to invest in and here’s what it said.” Please do not do that. ChatGPT is not capable of determining which stock is better than others, because that requires careful analysis and thought. What it is able to do, however, is answer your question anyway with extreme confidence.
About ChatGPT and Hallucinations
Because AI is not capable of thinking, it also does not have a concept of “knowing” versus “not knowing.” As an algorithm, ChatGPT will always determine what the next word should be, regardless of whether the answer is in its database.
As a result, sometimes if you ask ChatGPT something, it will give you an answer that is comically wrong. One of the most famous prompts is “are there any countries in Africa that start with K?” and it used to answer, with incredible confidence:
“While there are 54 recognized countries in Africa, none of them begin with the letter “K”. The closest is Kenya, which starts with a “K” sound, but is actually spelled with a “K” sound. It’s always interesting to learn new trivia facts like this.”
When Google released its own AI, it told people that putting glue on pizza would help with its taste and chewiness.
These are called “hallucinations,” although that word is somewhat offensive, as that is not what they are. Instead, the best way to understand what these are is that ChatGPT’s algorithm has to figure out what the next word may be, and if it doesn’t have an answer in its database, it will come up with the next closest one that it can find mathematically to finish the answer. It will also do so extremely confidently, because it doesn’t have the concept that it’s talking nonsense.
As a result of hallucinations, ChatGPT answers can be wrong, but presented in such a way that they appear to be confident with their answer.
Using AI With Your Work
I intended to do many more posts about AI in the future, as well as provide recommendations for how to use it, discussions about ethics, and more. There are other ways to use AI and other times where you should avoid using it, even when it appears to be a good use case.
For example, you should likely not use it to complete your marketing materials (we’ll talk about why at a later date), but you may find it very useful to quickly summarize a larger thought in a message or help you with social posts.
I find that one of the best AI tools for my purposes is a tool that provides meeting summaries of my Zoom meetings. It “listens” to the context of the meeting and them provides what are usually very accurate meeting notes to help us all remember what we chatted about.
AI tools are here, and many of them have incredible capabilities. But it is still important to understand what AI is and what it isn’t, so that we can understand how to use it now and in the future.
If you’re looking for more help starting your practice and want to better understand the tools that are available, please sign up for your kick starter consultation, today!