There has been a lot of furore on ChatGPT and its implications for the tech industry. So I decided to explore it myself. In this post, I will summarize my experience with ChatGPT.
Ready! Steady! Install
The installation was super cool. I was able to install ChatGPT without reading any documentation and IMHO that’s how all tools must be so that the experience is seamless. So if you want to install ChatGPT, just go to openai.com and click on the ‘try’ button. It will take you to the signup page. You can signup using email or you can use your Google or Microsoft logins to authenticate.
Once the authentication is complete, it will take you to the ChatGPTs landing page where you can do an AMA(ask me anything) to ChatGPT.
The coding Prompts!
Being a developer, the first thing I wanted to check is if ChatGpt can actually write code. So I went ahead and asked ChatGpt to write a code to connect to AWS. BOOM!! It started writing right away. I was curious to see which segment of AWS it will target and it targeted the SImpleStorageService a.k.a S3 to showcase a connectivity snippet to AWS. Below is the response I got:
It gave me a snippet right from the Boto3 documentation which prints the buckets from the account it has instantiated the client with. I copied the snippet and ran it in my IDE and it gave me the correct output as shown below:
Yes, it assumes that we know a lot of things but I was happy with the result as I got a working snippet. ChatGPT also gave me the explanation for the above snippet, the prerequisites requires to run the snippet and a link to set up AWS CLI using access keys.
Now I was curious to see what Google would return for the same query and I got the below result:
There were links to tutorials, forums, and getting started guides but none of them gave me a snippet that I could reverse engineer. Now I kind of realize why ChatGpt was seen as code red for many search engines, it doesn’t present you with choices unless required and it gives you a solution out of the box.
The Complex Prompt that was flagged!
Now I decided to try something little more complex and wanted to check how ChatGpt would respond to it. I was trying to migrate users from an AWS Cognito user pool and I asked ChatGPT to write a code for it. It gave me a response but flagged it as well!!!
It stated that my query violated its content policy and at the same time asked me to report if this was an error. As this is a normal use case where we export users to a new region or new account on account of Disaster recovery or High availability. I did try out the code in my IDE using a user-pool id and it didn’t work out of the box. I had to make a minor tweak and it worked.
The failed Identity prompt!
Now that ChatGpt was doing quite well in chalking out snippets of code, I wanted to explore how it performs when prompted with non-tech prompts. So I asked ChatGpt about myself and this is the response I got:
Well, the search engines are a clear winner here !
Phew! Just 2 prompts to ChatGpt and I have realized how potent this AI-powered application is. Watch out for this space for more findings on ChatGPT. It’s fun, it’s interesting, and it’s sparkling new!!