Skip to main content

Some thoughts on xeokit & AI

· 3 min read
Diyan Rashevski
Diyan Rashevski
COO @ Creoox

I've started an experiment with AI. It reignited the discussion about our position on AI within the xeokit team. Do we use and accept AI? For what? How?

Below are my personal thoughts on the topic.

Experiments vs. Responsibility​

I believe that we in the xeokit team have a very strong sense of duty and responsibility towards the xeokit community of integrators and users. So, out of fear of overpromising or letting someone down, experiments very often stay internal only.

The benefit of this approach is clear - whatever is there can be relied upon. The downside is not sharing that innovation spark, impulse and inspiration for someone to try something different.

Reaching a better balance typically goes through transparency. One such action was to start the xeoMeetup as a safe space for sharing. But it's not enough.

The elephant in the room - AI​

An elephant drinking tea at the Robur Tea Room in Sydney in 1939

Photo: “Elephant’s tea party, Robur Tea Room, Sydney, 24 March 1939,” by Sam Hood. Public domain. Source

AI is a very good fit for experimenting and this is what I did. Very quickly I was able to put together a few examples with xeokit that I had in mind for quite some time. In just a few hours. While having pizza.

I did not write any code for these examples myself and I did not review it line by line.

These examples are different from our typical xeokit examples. They are more complex and partially too long for a human to fully read and digest. Still, they bring value by demonstrating more complex capabilities with ideas on how they could be implemented.

xeokit is made by engineers for engineers

We stand behind that statement. AI is just a tool, a very powerful one. And with great power comes great responsibility...

There always has to be a human in the loop. Every part of the core xeokit is carefully tested and maintained by our team.

Finding the right balance with the examples​

The examples demonstrate how someone can use xeokit in different cases - customizing it and building different features. An example is often a starting point for developing a feature. It can also serve as an inspiration on how to design a feature.

The examples are not part of the core xeokit codebase. Our quality standards weren't lowered.

In this case, I was the human in the loop as the driving force and the team provided feedback. I specified, tested and finally approved the results.

Transparency, trust and freedom of choice​

This is something we strongly believe in and follow. So, also in this case, it was important for us to clearly indicate when an example was AI-generated.

The AI-generated badge can be off-putting for some and appealing for others. And both opinions are valid, especially considering different situations and requirements - from a developer building a heavily used worldwide application to an architect, BIM manager or civil engineer creating internal tools for their company.

The important thing is that you know what to expect when you see it.

Open Source & AI go hand in hand​

There are different ways to think about the impact of AI on open source. One visible trend are the exploding AI-generated contributions, which can overwhelm reviewers and maintainers.

On the other hand, AI also doesn't reinvent the wheel. Adoption and trust often come from building on proven technologies and AI makes it so much easier, even the default to use open source.