Exploring the potential of ChatGPT: how it could change the nature of work

While everyone was losing their mind over AI generated artwork this year I think in retrospect we'll look back and appreciate that AI text was the real MVP. Here are some of my ramblings on how this might affect our professional lives and how I used it to write some Salesforce code for me.

For better or worse AI will probably be hooked into everything we interact with at some point armed with contextual data and sensors. Until that slightly concerning day comes to pass ChatGPT still requires a text prompt in 2022. I classify these prompts roughly into two groups; are you providing all of the data needed to create an output (such as a private email for work) or are you asking it to summarize data (write me an essay about the civil war for my 8th grade history class).

Realistically this is somewhat of a a false dichotomy but I want to draw your focus to the concept that for lots of professional applications where you may be dealing with proprietary information the prompt will have to contain the majority of what you want to say.

So for some the benefits will fall under the marginal category. There are certainly exceptions such as when the input stays static but the output stays needs to change- ie: corporate twitter accounts and LinkedIn posts.

But do you know what happens millions of times every day, requires perfect formatting, and an encyclopedia of knowledge? Programming. ChatGPT incredible for automating the everyday building blocks of Programming.

To test I wanted to challenge ChatGPT to write something that...

1. Was proprietary (to Salesforce) but well documented publicly
2. Required multiple scripts that all needed to relate together (in this case 3)
3. Called an external API for information

So I prompted it to write a re-usable web UX component that displays the temperature.

This initial answer blew my mind. It understood what I wanted. It found the NWS API reference. It wrote perfect code for Salesforce. It happened instantly.

Although- this first task also illustrates one of ChatGPT's greatest weaknesses; the combination of 100% confidence while providing blatantly incorrect information. That NWS API URL it provided DOESN'T ACTUALLY PROVIDE THE TEMPERATURE. Okay I'm being a little sensational but I do think this particular combination of personality traits can create some ethical issues in other areas.

If you're a pessimist you might be thinking "Okay- so this is just a slight upgrade on Stack Overflow?". Here is where the 'Chat' part of 'ChatGPT' starts to show its strength.

"How is this so good?" I kept asking myself. With a little direction (me clicking a the original end point and not seeing temperature) ChatGPT can dig a layer deeper, get the correct answer, and then update the code.

And the same goes for ALL the other road bumps I encountered like this compiling error I received when trying to commit the output to Salesforce.

ChatGPT finished up by writing the Aura component and the JS controller

In summary OpenAI has created something amazing and made me feel like a kid again. If you have any interest I urge you to check it out now. It's currently free but CEO Sam Altman recently mentioned that "we will have to monetize it somehow at some point; the compute costs are eye-watering" so don't count on this

And just in case you were wondering- I wrote this entire article keystroke by keystroke. But don't assume that everything you read has the human touch.

#chatgpt #gpt3 #ai #salesengineering #salesforce

The views expressed in this article are my own and do not necessarily reflect the views of my employer