Like most of you, I’ve been intrigued by the advancements in Artificial Intelligence (AI) over the past year. The public unveiling of ChatGPT at the end of 2022 brought the world of generative AI to our virtual doorsteps. Reactions varied wildly. Some folks were blown away by the seemingly endless potential for AI to make our jobs easier. Others were worried that this new AI would make their jobs obsolete. Some highly influential tech-leaders even expressed serious concerns that the new developments in the technology could pose existential threats to humanity.
I can’t tell you which scenario is true for any or all of us. But I can share a little advice from a business advisor’s perspective. In short, technology like OpenAI’s ChatGPT can be a highly effective way to speed up the content creation aspect of your work (e.g., crafting social media posts, giving you good ideas for a blog post, helping you proofread your emails). However, Large Language Models (LLMs), like those serving as the technology behind ChatGPT, are prone to error. Sometimes they hallucinate, meaning it generates text that isn’t factually correct or is entirely fabricated. For example, an acquaintance of mine (Michael) told me about using ChatGPT to summarize information about a potential business partner ahead of a meeting. On the call, Michael asked the man about a particular research center he had created as part of his work. The man was confused. He had never heard of the center, nor had his company ever created a research center as part of their work. ChatGPT had invented the entire thing! Overall, LLMs like the ones featured in OpenAI’s ChatGPT and Google’s Bard seem to be fairly accurate and reliable, but you should always use these resources with a buyer-beware mindset, especially when accuracy is imperative (e.g., asking ChatGPT to review a contract or other legal document).
When it comes to your finances, though, you really should leave this to human professionals and industry-tested technologies. A major reason for this is that although it may look like LLMs can solve math problems for you, they aren’t doing it like you think they are. In over-simplifying terms, LLMs are always trying to predict the next word in a sequence. Say you ask ChatGPT “What is my anticipated return on $19,000 invested at a 6.5% continuously-compounding rate if I leave it alone for 14 months?” The LLM is doing some math to get your response, but it’s not the math you think it is doing.
To generate its response the models are predicting the words that have the highest probability of being seen after the words you presented. You can try part of this yourself. For example, I would predict that if you asked a financial adviser that question, the first words they would say in response would be “If you invested $19,000 at 6.5%...” This seems like a pretty normal and expected start to a response. But what comes next? The LLMs in the chatbot are busy at work deciding which word, out of all the words we know in the English language, are most likely to follow.
It’s entirely possible that the chatbot you use will return the correct answer. But if it does, it will be more of a case of “I knew you were going to say that was the answer!” as opposed to “I input your numbers into an equation and did the math.”
What does all of this mean for you? It means that you should have fun exploring the ways that generative AI like LLMs can help you with your daily workflow. Some aspects of your work very well may be improved by a resource like ChatGPT. But when it comes to high stakes information and requests, it’s best to follow tested and verified ways of getting things right.
ChatGPT may not be the solution you are looking for, but when it comes to your financial, payroll, and accounting needs, I’m always here to help. Just reach out and we can discuss the solutions that are right for you and your business.
Comments