Hello World! Yesterday, Feb 21, 2023, I got an opportunity to Explore AI tools with Alexandra Martinez - a developer advocate at MuleSoft. It was my first time doing a live stream, and a pretty fun experience. Thanks to all who came out to support and provide valuable prompts.
We explored six AI tools; however, this post is strictly a recap of our conversation with ChatGPT Plus about MuleSoft's history. This is constructed with two tests, screenshots of related prompts and answers, followed by my observations.
The purpose of the chosen prompts is to first find out what ChatGPT Plus knows about MuleSoft and then access its accuracy.
Posts about our conversation with ChatGPT Plus about Mule 4 development, DataWeave, and Connectors will be available in the subsequent weeks.
Note: Interested in the ChatGPT MuleSoft and Mule 4 conversation? You can access it here. If you are using a screen reader, I do not place the actual response in text form on this post. You will need to access the link to get the ChatGPT response.
What is ChatGPT Plus?
ChatGPT Plus is a subscription-based conversational chatbot tool powered by Generative AI. It was created by an artificial intelligence company called Open AI. The research version of ChatGPT is free. ChatGPT Plus subscribers can choose between the Default (latest version) and Legacy (Research version) model.
Let's move on. So let's see how much was ChatGPT trained on MuleSoft and its history.
Test 1: What does ChatGPT know about MuleSoft’s History?
We asked the following questions
What is MuleSoft?
When was it founded?
Did MuleSoft have a co-founder?
Observation
The screenshots above show that the ChatGPT plus default model was trained on MuleSoft’s history. The bot spoke in a confident tone and appeared very knowledgeable about MuleSoft, from a startup to becoming a Salesforce acquisition. During this test, I accepted ChatGPT's response as fact and did not do any fact-checking. The confident tone does make this ChatBot extremely convincing.
I feel like ChatGPT is my super intelligent smart best friend. - Whitney Akinola
Spoiler Alert: Keep reading.
Test 2: What does ChatGPT know about MuleSoft acquisitions?
We asked the following questions
Who did MuleSoft acquire?
Who acquired Servicetrace?
Observation
This is where ChatGPT Plus gets the answer wrong. Let’s look at the acquisitions listed.
ProgrammableWeb - This is true. Unfortunately, I found this fact this year after MuleSoft announced the site’s requirement. ProgrammableWeb had a great run and served its purpose.
CloudHub - To my knowledge, MuleSoft has always owned CloudHub. I could not find an acquisition. However, I started to doubt myself due to ChatGPT Plus’ confident tone. Let me know in the comments below if this is accurate.
Velostrata - This is misinformation. Velostrata created a cloud migration technology and was acquired by Google in 2018.
Lightwell - This also appears to be misinformation. Lightwell was acquired by Eliassen Group in 2021.
But what about MuleSoft’s (by Salesforce) Servicetrace acquisition?
The chatbot does not know about the Servicetrace acquisition because it was not trained. The chatbot mentioned the default model only knows up to September 2021. Servicetrace was acquired by MuleSoft in October 2021. This is a limitation. The world is constantly changing, and I guess that’s the limitation of a pre-trained model. I wonder how a model will be that is trained in real time using Bing search or Google search will work.
Note: MicroSoft has released Bing with the ChatGPT integration in a limited preview. Google is testing its ChatBot tool pre-trained using the laMDA model name, BARD. Style points Google on choosing BARD as the name.
Missed the Twitch Live Stream?
Click Watch Video
My Verdict
I thoroughly enjoyed conversing with ChatGPT Plus. I always get interesting information from this Chatbot. As of February 21, 2023, ChatGPT Plus is inaccurate concerning MuleSoft history. However, I am still optimistic that as users adopt this new tool and start to provide feedback on misinformation and bias, this chatbot will only get better. I encourage you to give it a try. You can try the research version for free.
What are your thoughts? To keep this post short and to the point, I will release subsequent posts on other prompts that came up during the stream.
Be sure to subscribe to my blog to avoid missing a post.
It's great that you created a blog post from your findings! It was very insightful :)