EA and ChatGPT 5 - the death of the forum?

[Ian Mitchell]

This forum has saved me on many occasions, and is beyond question the definitive source of wisdom about the use of EA.
But it's days as the interface into that wisdom may be ending.
Over the last few months I have been using ChatGPT5 to answer my questions about the depths of EA.
And the results are outstanding. (this may apply to other LLMs - I haven't tried any others).
The answers it has given me are at least on the same level as 'Inside EA', may be even more detailed. Almost Geert-level brilliance.
OpenAi have clearly scanned all of the posts in this forum, to the extent that it will even, occasionally, say WHO provided an answer, which is the proof of the source.
Scanning material like this may not be a Good Thing from an Intellectual Property point of view, but it's definitely a thing, and it seems sensible to use it.
Which is where the issue lies: if we all use an LLM to find out about EA, then where does that leave this forum? Why ask the forum, when ChatGPT has all the answers? Where does it leave ALL forums? Does OpenAI just continue to absorb all our collective knowledge, then charge us to use it?
I'm not suggesting an answer here, just pointing out what the future might look like.
I look forward to hearing other people's experiences of LLMs and EA.

[Geert Bellekens]

I see that as a good thing. AI is sort-of an improved search engine.
It avoids me having to answer the same question here over and over again.

For new questions, I'm guessing we still need something like the forum.

Geert

PS. It can also be a good marketing tool. I once asked it to recommend a Sparx consultant for things like scripting and document generation, and I was very pleased that it recommended me

[Takeshi K]

Hi Ian,
I'm glad to see your name here again. I hope to see you in person again soon.

In any case, I definitely agree that LLMs are extremely powerful and have changed many things.
Regarding this forum, I think LLMs are a very user-friendly UI, but that's all. The current search feature on this forum is not very effective, and the results contain lots of irrelevant and/or unwanted posts. Using LLMs means we don't need to search for the expected posts.
(I must say that this does not imply any blame for the search feature or this forum. Current search features including other forum softwre are all very basic and not very clever.

However, LLMs only know the information that has already been posted. There are and will be many unpublished questions and tips. We (human) understand the situation, requests and constraints, and we offer the best solution for every user.
(Of course, if LLMs were much better, they could provide the same or similar information as us...)

[PDC]

OpenAi have clearly scanned all of the posts in this forum, to the extent that it will even, occasionally, say WHO provided an answer, which is the proof of the source.
Scanning material like this may not be a Good Thing from an Intellectual Property point of view, but it's definitely a thing, and it seems sensible to use it.
Which is where the issue lies: if we all use an LLM to find out about EA, then where does that leave this forum? Why ask the forum, when ChatGPT has all the answers?

AI is sort-of an improved search engine.
It avoids me having to answer the same question here over and over again.
For new questions, I'm guessing we still need something like the forum.

...LLMs only know the information that has already been posted. There are and will be many unpublished questions and tips. We (human) understand the situation, requests and constraints, and we offer the best solution for every user.

For me, this sums it up well. The LLMs are not (yet) modelling tools, and whilst they will/should learn from their own experiences, they cannot be 'creative' in the human sense of bringing meaning to the UML/SysML/EA/etc semantics in a way that can be interpreted by a large number of users, or in a very specific scenario.

As you suggested, yes maybe this makes forum replies simply feeders into the LLM... but I don't think it necessarily cuts off the source.

Of course, as a systems engineer and architect I have a vested interest in ensuring I still have a relevant career and am not replaced by AI in the next 12 months. So I'm always going to argue strongly FOR humans as modellers and supportive communities!

[adepreter]

If nobody adds innovative knowledge to the forum, I wonder what the impact will be on AI's ability to collect a decent level of new and valuable experience-based knowledge.

[Sunshine]

AI needs constant feed of information to be input. Just suppose we all stop using this forum when the next version of Sparx EA comes out then the usefulness of AI will degrade as there will be no new information on the next version of Sparx EA that the users have discovered other than what is in the manuals.

[MattAtRamsay]

As you interact with AI and its current BoK, and work through a problem to resolution, it learns that also.

[ZuluTen]

Up until now, all my EA scripting has been done in VB, but with the advent of Scriptlets I thought that I should move to JavaScript, and in my laziness, and prompted by Ian's post I thought "let's ask Chat GPT to write some of the code..."

The results are fascinating: it has quite confidently generated the code, but included lines such as this:

var diags = Repository.GetDiagramsByName(name);

....which would be admirable if the Repository class actually provided a "GetDiagramsByName" method. Unfortunately there is only GetDiagramByID and GetDiagramsByGUID, so it seems that ChatGPT has simply invented its own method!

It was quick to apologise once I pointed out, but doesn't inspire confidence.

[Sunshine]

I've similarly asked, AI to help with javascripts and never once has one it generated code that worked without modification, often AI hallucinated non-existent API calls. That being said it has reduced time to write code from hours to abot 20 to 30 mins

[Geert Bellekens]

I've had the same experience. It's a great timesaver, if you know what the result should look like.

If you have no clue, it might be very time consuming to figure out where exactly it goes astray.

Geert

[Modesto Vega]

I've had the same experience. It's a great timesaver, if you know what the result should look like.

If you have no clue, it might be very time consuming to figure out where exactly it goes astray.

Geert

This seems to be the problem with all AI tools, if you know what you are doing or what you are expecting, they can be a good productivity tool. If you don't they just generate rubbish.

[Paolo F Cantoni]

I've had the same experience. It's a great timesaver if you know what the result should look like.

If you have no clue, it might be very time-consuming to figure out where exactly it goes astray.

Geert

This seems to be the problem with all AI tools: if you know what you are doing or what you are expecting, they can be a good productivity tool.  If you don’t, they just generate rubbish.

(my emphasis)
Not quite Modesto...
They MAY generate rubbish. The problem is that if you don't know what you’re doing, you can't tell if it is rubbish!  This is worse than consistently creating rubbish.

Paolo