January 4, 2023 | 3 min read
Chatting with ChatGPT about ChatGPT
Mario DiBenedetto
Managing Director
Chances are, you’ve seen the multitude of posts associated with the latest “chat bot” technology called ChatGPT. What makes it particularly interesting is that it is “trained” on a vast amount of information such that it can actually write most of this article. Except for my prompts and explainers, I am going to try to illustrate to power, potential, and limitations of this new technology by using this new technology. Let’s start with…
We need a little more detail than that....
Ok, so now we know it's throrough....but is it self-aware?
Wow! There is some really important information in that response! And it’s much better than a salesperson who promises technology that can solve every problem on Earth!
So let’s start getting specific to our business needs…
Well, to be honest, that sounds just like the chat bots that are available today. Am I wrong?
So, ChatAI has already been trained on a specific dataset of text and that makes it able to act as a generalist to provide coherent responses to questions. Which means that “out of the box”, it could provide things like borrower education because that is a relatively static set of knowledge that already exists on the public internet. That doesn’t make it particularly useful for business-specific use cases where the underlying answers might undergo change.
OK, that opens up a lot more opportunities! Imagine training a bot simply by feeding it your underwriting guidelines. Could it feed you answers to borrower eligibility questions? Would it do so better than your pricing engine? What if, instead of YOU training the BOT, your investors trained it (and maintained it!)? Could it produce the decision condition list on a loan if you fed it the data? Maybe. But, let’s remember that the answer to our second question was that this technology is not good for “making decisions or taking actions”.
So is this just a better chat bot and nothing more?
