Spatial Awareness for trinity 3.2
AI is the buzzword of the moment. ChatGPT has brought it into the mainstream.
I think we are all wondering how it will effect our lives, and more specifically our jobs.
We've long been tracking this programmatical approach to problem solving and its time to share some results...
We are releasing this in the form of a plugin for trinity 3.2 and above, dubbed Spatial Awareness.
The plugin will be free to existing subscribers, although won't be part of the standard system. Why so? The plugin alone is of little use without the training data, and at the moment this requires some considerable offline work on our part.
After working with brainjs and subsequently tensorflow, our experience has been that whilst powerful, these libraries, and AI in general is extremely inefficient. We could upgrade to cloud based GPU servers for a purely AI, brute force approach, but that goes against our ethos of providing affordable, elegant solutions, not to mention the environmental impact. Instead we developed a hybrid approach, marrying the magic of AI with the speed of more old school 3D mathematics.
Our first model is based on wine rack data. There were a few reasons for this.
We delivered our first wine cellar designer almost a decade ago, so we have access to masses of real world designs/data.
Designing the wall of a wine cellar is a pretty empirical process compared to designing a living room (or writing a poem...). Whilst there is room for "imagination", the answer is very clearly either right or wrong.
This brings us to our brief road map for the future.
Having Spatial Awareness is cool and everything, but what about Colour Awareness, what about Style Awareness? These are essential elements for an Interior designer, and therewith essential ingredients for a system aimed at helping the novice achieve aesthetically pleasing and functional results.
The good news is they are in the pipeline, and if you want to be part of the beta program, it's as easy as sending us an email. Though be aware that training a data model takes an awful lot of time. We can provide tools and guidance for you to help us do this, but can't always guarantee instantly usable results. This should change over time as we keep polishing our algorithms and finding new ways to extract patterns from existing data.