Graphical Industrial Design using Bing AI
In this previous post I introduced the idea of qualitative programming, i.e. of developing programs that can process qualitative data natively (instead of having to convert qualitative data to digital form, process the digital data, and then convert the result back to a useful qualitative representation for the end user).
The following information process was suggested:
- Accept human input via a UI.
- Perform qualitative analysis of the human input.
- Process the discrete (analysed) qualitative components of the input.
- Reassemble the qualitative program output into a material, holistic form.
- Present the output to the user.
In this post, I will attempt to apply this method to graphical industrial design.
Choosing a Problem
Interior Design is a relatively easy problem for an AI image processor. This is because the graphical design consists of the placement of discrete elements, e.g. furniture and plants, within an image (as opposed to the holistic design of a single artifact, such as a car or a watch). So I have chosen Interior Design as the subject of this experiment.
Applying the above method should allow application of the following AI-driven design process.
| Qualitative Programming Step | Application |
| Accept human input via a UI. | Allow input of an Interior Design image / textual description. |
| Perform qualitative analysis of the human input, in human terms. | Analyse the moods invoked by the design. |
| Process the discrete qualitative aspects of the program input. | Adjust the design to suit the required mood. |
| Reassemble the qualitative program output into a holistic form. | Combine the unchanged and adjusted parts of the Interior Design layout. |
| Present the output to the user. | Present a final Interior Design image to the user. |
Qualitative Vectors of Interior Design
In this post I introduced the idea that any system can be defined using three basic qualities:
- Logical component
- Control component
- Energy component
Applying this qualitative analysis paradigm to Interior Design gives approximately:
| Qualitative Component | Equivalent in Interior Design |
| Logical | Functional aspect |
| Control | Austerity aspect |
| Energy | Fun vibe aspect |
This means that the qualitative aspects of any interior design can be defined in terms of its functional aspect, austerity aspect and fun, vibe aspect.
Uncontrolled Output
Let’s ask Bing AI to create
- a functional design for a living room,
- an austere design for a living room and
- a fun, lively design for a living room.
These designs are extreme and not very useful, let’s see if we can get Bing AI to generate some more subtle living room layouts.
Qualitatively Modifying an Existing Interior Design
Instead of asking Bing AI to create a layout for a living room from scratch, let’s ask it to modify an existing design, by enhancing the functional, austerity, and fun aspects of the existing design.
The base living room design is this image:
Qualitatively modifying this design gives the following results.
Together, these 12 images show subtle variations on a theme of the basic original design, and provide useful food for thought when considering an actual living room layout.
Combining Designs
Let’s say we want to combine two living room layouts and see what the result would look like. There are two ways we can do this; we can either just ask Bing AI to combine the images and see what it comes up with, or we can specify which aspects of the designs we want Bing AI to combine. This second option gives us more control over the results and allows us to experiment with different variations of combinations.
One of the images used in this experiment is the layout shown above, the other is this living room design:
First, let’s try a simple combination:
Now let’s try combining specific aspects of the two designs.
This gives the following results.
In this case, it is not clear which aspects of the images have been kept and merged, but this text suggestion did produce different results than the original simple merge request, and therefore could be modified to produce variations on a theme of the original designs.
Conclusion
Bing AI as a general purpose tool is probably not quite there yet in terms of extracting specific qualitative aspects of Interior Decoration designs and recombining them, but it is an interesting tool that is capable of generating that exhibit different moods and nuances and that can serve as food for thought for actual designs.
Presumably future versions of Bing AI will possess more powerful feature extraction and recombination faculties that will make it a more suitable tool for generating mood-driven industrial designs.
Note:
- All images in this post were generated with Bing AI chat set to “More Precise”.
- If Bing AI goes on strike and suddenly tells you that it is incapable of generating images, close the browser tab and try again.