AI VS Access database

[mobarak ahmed]

Hello everyone,

With the rapid advancement of AI, are there any platforms or applications where you can simply describe the tables, queries, forms, and reports you need, and it will generate an Access database to meet those requirements?

 

[The_Doc_Man]

One problem is that various attempts to create database technology without actually going through a detailed design phase usually led to a poorer product than one designed in detail manually. For a while, PeopleSoft (now a wholly owned subsidiary of ORACLE Corp) tried doing something similar to that idea, with some level of success but with some problems to go with it.

Their problem was that AI wasn't nearly as advanced when they started the product and so they used a "library" concept. You could pick and choose generic data groups for persons, jobs, inventory, etc. Their product was data-driven. More precisely, they had the equivalent of pre-defined class objects for a person, a job, an inventory item, etc. that contained data and the code to process same.

The problem was soon revealed that if you had ANY anomalies in your data, you would require manual intervention. If you had ANY unusual actions that were needed, their processing library needed manual intervention and support for the odd procedures. It also had high overhead loads because much of what they did was based on interpretive scripts at many levels. The Navy tried using PeopleSoft in the abortive DIMHRS project, which eventually got scrapped because it was a resource hog and required a lot more (very expensive) customization than was originally anticipated - or budgeted.

I would also be wary of AI "hallucinations" - which is the term used to describe what you get when the AI goes off-topic on you. So far, no one knows why it does so, but many AI platforms do appear to "make up" stuff that isn't relevant to the question. These hallucinations have begun to be documented. You can search for "AI hallucination" to see articles on that subject.

I would say that if you were going to use an AI solution, go ahead and get your answer in the ways suggested by Uncle Gizmo, but don't blindly trust ANY answer. Examine it closely to verify that it actually makes sense.

 

[GPGeorge]

Hello everyone,

With the rapid advancement of AI, are there any platforms or applications where you can simply describe the tables, queries, forms, and reports you need, and it will generate an Access database to meet those requirements?

I have been participating in a discussion on a different forum where the OP posted a solution created with the assistance of CoPilot.

Copilot, Access, and VBA | Microsoft Community Hub

I don't know how many of you use Copilot to assist you with Access. I developed my own application using Copilot and it was extremely helpful. I...

techcommunity.microsoft.com

My thoughts on that project are included in the discussion.

 

[GPGeorge]

Hi! Yes, tools like Microsoft Power Apps and OpenAI's Codex can help generate database components based on descriptions. They’re not fully automated for complex Access databases yet, but they’re a great starting point!

Microsoft Power Apps? Generate database components?

I believe you may be thinking of Copilot for PowerApps in the case of Microsoft. And it relies on the internal models already defined in Dataverse. See these screenshots showing the use of "generative AI" and Copilot and drawing from pre-defined models, such as "Track sales leads". This might get a novice off the starting line, but they immediately run into the issue of what to do at the second step, where they have to decide whether the pre-defined model is appropriate for their needs.

The resulting tables are going to be in Dataverse anyway. And that means they are next to useless for any serious Access database application.

I am a fan of PowerApps, but not so much Dataverse.

 

[Cotswold]

I just wonder about using AI to create a serious database application at this point in time. Maybe AI can make a reasonable attempt? I don't know. We all will have had to create our own program generator for Access, or another language used. Just to avoid wasting time typing mainly the same code for every screen. Why go to the bother of placing often the same buttons onto every screen, although not always the same position, together with their events when a small UDF will do it for you effortlessly?

If an application was generated from a specification by AI, how easy would it be to update or amend it in a year or two? As I understand it AI is an evolving environment. So, if it generated amendments to a program later, would they be correctly applied to the old existing code in use?

Exactly where would you be if the amendments caused issues in earlier code? Could you sort that quickly? Or would you have to create a whole new specification with the changes and hope it will handle existing data? The AI code will be written in a style not your own, with strange variable names. If you don't like the resulting screens, or reports I presume you will need to change them manually. Could the AI understand your updates and manage them if you needed it to create changes?

There is no doubt that many companies will want to use AI to reduce software costs. When that ambition becomes reality I cannot guess. A bit like all those who thought they'd put something together with a spreadsheet but of course in reality never could.

For some reason that escapes me, billions have been spent on driverless cars but without success. But maybe in 20 years they'll manage it? Unless you can get in and say “get me there in less than an hour, I’m late” what’s the point?

Mind you there are many businesses that employ experienced people to make decisions and complete specific tasks. Solicitors, property management, accountants, estate agents, the tax man, the local doctor etc. None of them are doing anything special and can easily be replaced with software. You are in the main, only paying them for their knowledge. Local authorities and huge swathes of government can be removed. They only add cost and are in the main totally unskilled and make poor decisions. In theory all could easily be replaced with a computer running AI that will not forget anything it was taught. Maybe tell your children and grandchildren to learn to lay bricks, be a plumber, or a joiner? Big saving on university fees.

AI like everything else in the early days you are given the impression that it is all about to happen. The solar panel was developed in the 1960s but took over 40 years to be in common use. The internet has changed the way we do things but up to now not what we do. You may not guess this but at the moment I'm a bit sceptical about AI and software development. Can that manual work actually be computerised? I'm not sure.

 

[AccessBlaster]

Where AI will shine in database construction is in table design IMO. For instance, let's say you need a table design for an inventory project. AI could help design a table(s) structure that includes normalized to N3 with all PK and FK related. It would be just a blue print, you would still need to build it and understand a fair percentage of the design.

Having a complex table design knocked out in a matter of hours not days or weeks helps with every aspect of the project and would save time and money.

 

[AccessBlaster]

I'm not so sure that an AI could create a schema unless there were models to help it. And the reason is the same reason that humans have so much trouble. It takes a lot of work to interview users and review existing paperwork to identify related items and group them into tables. Then to group the tables to define relationships. Things like repeating groups are difficult too. Groups like month names and day names become obvious because the value list is fixed but what about things like expense types? Electricity, Gas, Oil, Snow Removal, Lawn Mowing, etc. This list is almost infinite depending on what your main object it.

The reason that more experienced developers are better at this stuff than new ones is because they have more general life experience and have interacted with many more types of computer systems over time and are less inclined to believe users who tell them that there will NEVER be more than two roommates in a single dorm room.

I agree 100%. What I know from playing around with AI is, it's conversational. While the first few drafts of your project will be wonky, AI remembers the mission and makes necessary tweaks at our prompting. It will not be right for all things, my guess is, it will be very good at design.

 

[AccessBlaster]

It doesn't actually (at this time) have the ability to come up with something new.

Correct, it analyzes the available information from real time online sources and makes a judgement about your prompt request. It's just another tool within a vast array of tools.