The Next Thing I Want to Do is Try and Extract Some Financial Data from Different Large Language Models
A recent financial data and see how they behave. So, let’s say I ask, ‘Can you tell me the weekly stock price of Google stock for the past one week?’ Actually, let’s get one data point first and then we’ll ask, ‘Can you tell me the latest stock price of Google stock?’ Let’s see if GPT can at least pull this off.
And then we go to Bing. I know Bing will do a decent job because when you ask it to fetch specific information, it can’t. So, it’s saying it doesn’t have the capability of providing real-time updates. I’m going to go again and fetch the data from the internet. Let’s see if that helps.
So, as of 10 AM EST on May 20th, the stock price of Alphabet was 122.76. Its market capitalization is this. So, of course, Google did a really good job at it. Now, can you tell me the price? Tell me the closing price for the past seven days in a table format. Again, GPT failed. It’s not able to search through the internet and get me the data that I’m looking for. 122.76. Bing did a good job at it.
Okay, now it got me the table. I’m going to run the same command in Bing just for us to be able to compare. Okay, so actually Bing went one step ahead and it got this data from MSN. It actually plotted the graph for us. That’s pretty amazing. So, I don’t need to ask my second prompt. It has already plotted it for us. Can you plot this in a graph? Let’s see if Bart can do an image or not. So, it wrote me a code snippet instead of actually plotting the graph. That means that it can give me the code to do the job but not the graph or the image itself yet. But it will definitely be included in Bart soon as well. GPT has not fetched the data for me, so I cannot ask it for any further questions. Bing was able to get me the graph. Let’s see if you can give us the price in a table format for the past seven days. I can’t find a table format for closing price because it’s actually fetching the information instead of creating it. That’s why the answer is different. So, in this way, you see although Bing and Bart can both fetch the data while web browsing plugin for OpenAI is broken right now and it’s not giving us such good information, Google and Bart again won and did the best job when it comes to understanding what we’re looking for and actually providing the information.
Let’s try one more thing. I would go to Bart and ask it to provide me the link to provide me a summary of the latest earnings report from Google. So, let’s see if we can fetch the return. Get me the latest earnings report. Google did a quarterly earnings call recently, so I’m expecting that the data should be there. Awesome! Here’s the summary of the latest earnings report from Google: revenues, earnings per share, Cloud Revenue, active users. Okay, what is the source of this data? Okay, let’s go check this out. What this link is. So, this is actually the earnings release file, a PDF file that Google released, and that is what the model is using here to fetch the information for us. Awesome! Great!
In the next lecture, let’s see if we can figure out if these models can work with external PDF links. That’s the next thing we’re going to try.
