Surely ChatGPT can't do Oracle as well…

Anyone that has not been living under a rock for the last few months can't have failed to notice that there’s only one tech game in town right now: ChatGPT. Every day the papers are full of stories of the thousands of jobs that will be replaced by the new king of AI—BT being a recent example. Universities and schools are grappling with how to stop wholesale (and undetectable) plagiarism and cheating. And Twitter and Facebook are full of endless ‘get rich quick’ schemes by which you can become an overnight millionaire with ChatGTP doing your legwork.

But surely there are limits… surely there will always be a bastion of expertise and know-how that will remain immune to ChatGPT?

I am very proud of the team of Oracle professionals I work with. They are some of the smartest and most experienced experts in the field of Oracle ERP that I have ever met. What they don't know about the dark arts of integrating, migrating or manipulating Oracle ERP data is not worth knowing.

So imagine how my heart sank when one of my team related the following story to me last week...

"I was playing with ChatGPT last night, and asked it to generate some EBS SQL code. It was pretty good…."

The following is what it produced:

For those not of a technical bias, this is an SQL statement to carry out the mass deletion of Oracle E-Business Suite (EBS) GL Journal Headers and Lines. The syntax is perfect, the pre and post execution steps are flawless, and this script would do an excellent job of purging those records. Typically I would expect an individual to have a significant level of experience in coding with Oracle EBS to be able to flawlessly produce such a script.

So why would I not simply fire them all tomorrow, and replace the team with low grade (and therefore far cheaper) developers overnight? If they could generate code using ChatGPT, why do I need expert knowledge and experience in my team?

But this is indeed a perfect example of the danger that tools like ChatGPT represent. If any individual were to execute this statement against their EBS instance, they would immediately open themselves up to a world of potential problems.

There may be nuances of this specific EBS instance that could lead to corruption or loss of referential integrity. The outcome of this script could have a devasting impact on the viability of that EBS instance—and the junior resource that applied it, based on the guidance of ChatGPT, would not know where to start to fix any of these issues.

And most importantly, deleting data directly from EBS base tables in this way could leave the organization in breach of the terms of their Oracle Support agreement—if any issues or problems arose in future, Oracle support would be perfectly entitled to say "sorry, you have corrupted your database, it's your problem". Imagine the impact that would have on an organization dependent on EBS.

ChatGPT is an inherently flawed tool that, by design, makes inferences and assumptions based on large data sets. But the answers are produced with such authority, that only an expert can find the flaws in the logic... It’s the AI equivalent of  "If you say something with sufficient confidence, people will believe you know what you’re talking about."

Do new AI developments such as ChatGPT represent tooling that can add significant value to the toolkit of a professional developer? Absolutely, and we should all embrace the productivity gains they can offer.

Are they capable of removing the need for experience, expertise and all-round nous that professional developers like those within the Inoapps team can offer? Absolutely not. 

Will that change over time? Perhaps, but right now there remains a place for the experts.