Gad Tan

An Inconvenient Question

Update: This article was originally published on 23 September 2015 and updated on 5 May 2017. As of 16 January 2017, it turns out that the "AI" portion of thegrid isn't really "intelligent" at all. In fact, the precomposed templates paying members could select from were criticised for being boring, expected and unintuitive.

Last Sunday, I received a text from Yong:

Automated Design or Machine-Aided Design has been something that's been gripping social media lately. Maybe this fascination and fear started because more people caught wind of this. If content-centric and user-specific design can be applied to web sites, surely it can be applied to branding can't it? The short answer is yes, of course it can. How to achieve autonomous design and to what degree of autonomy, is an entirely different question.

1. Machine-aided design, the design partner

Perhaps one of the most widely known uses of machine-partnered design is the mobile (and now desktop) application Flipboard. In 2010, Flipboard developed a layout system called Pages. Pages produced beautiful layouts, effectively replicated the respective brand identities and typography of originating publications. It utilised CSS3, SVG and JavaScript to make rendering as high fidelity and snappy as possible on early mobile devices (i.e. the original iPad running iOS 3+). The footprint for a publication averaged around 90kb for layouts, styling, fonts and images–lighter than the equivalent web page or a single photograph from an article. Sorta like an RSS feed.

A human designer was first tasked to create a set of 20 odd page layouts, divided up into 768px portrait and 1024px landscape orientations. From this set of predesigned layouts, Pages selected the layout that best fits the desired content, inserted the content into the chosen layout, and displayed a final page. With this template-based approach, Flipboard relied on designers to make layouts clear, distinct and beautiful. However, while Pages was able to generate from predesigned layouts, it only worked at a specifically designed size. New formats, operating systems and screen sizes presented a new challenge: fluid window sizes.

Following on the same vein of a machine partnering with a human, Flipboard later developed Duplo: a machine to partner with a machine—a full machine partnership. Loosely, Duplo is a new layout engine that starts with the ideas in Pages but uses a modular block and grid system to quickly fit content into thousands of page layouts in all sizes.

It starts in a similar way to the original Pages: A human designer creates a set of layouts. From how I understand it, these layouts are designed more as components that exist as sub-sets. From these sub-sets, Duplo looks at anywhere between 2,000 to 6,000 candidates, searching for the best layout to assemble and fit the content based on a set of guidelines.

Similarly, The Grid openly states on their website that base layouts were pre-composed by their in-house human designers.

The pattern we see in machine-aided / machine-partnered design is that a human hand has defined the parameters, contributing in the direction of the eventual output.

This opens up another facet of discussion: If different projects were defined with similar parameters, would it produce similar outcomes?

To digress, something that's closer to my daily craft is the criticism that sites these days "all look the same". Responsive Web Design (RWD) has often been accused as the culprit for causing sites to look too similar. Most designers these days are looking to craft a layout to fulfil a certain set of parameters (i.e. must be mobile responsive, must not have Flash) and as a result, produced sites that lack real thought. More often than not, the real culprit of this apparent similarity is a lack of consideration for the sites' contents and objectives.

In this case, the human has stopped thinking, and attempts to be a machine, reducing itself to checking a list of boxes. Thousands and thousands of cookie-cutter sites litter the internet, each one drawing just a little too much influence from the last. The access and proliferation of cheap, pre-built templates should not be blamed. Instead, the real malfunction lies in the profession of designers who claim to be building "bespoke" sites that might as well be machine-made.

2. Fully automated design, the replacement designer

Some of my contemporaries and fellow designers believe that nothing is original, and that everything is a remix of whatever we've seen before. Hence, for them an automated designer would be one that they would be able to input everything they have seen and experienced before into a machine and parse some arguments into it. From this, they would be able to generate whole branding schemes.

Yet, how different is this from today's common industry practice of hiring cheap labour (read: fresh student graduates) to produce formulaic design schemes for clients? Are designers today, merely automatons? If this is the definition, then replacing these cheap labourers would be fairly easy, and the future of branding/graphic/web design would be pretty bleak.

Perhaps the acceptance of this mode of operation starts during the education of a young designer. Quite frankly, I'm pretty disgusted by the curriculum that the design schools are adopting these days: Grab a bunch of references from FFFFound and Pinterest, change a few colours, fonts and slap on a new logo, bang, that's design, you're a creative. Branding? No problem. Just check out BP&O or the Behance portfolio of some hotshot studio and "adapt".

To me at least, the fundamental objective of branding is to create a unique association and image in the consumer or viewer's mind. Perhaps somewhere in a designer's career or education, the definition of unique got diluted and it became acceptable to just make some tweaks to someone else's work instead. Conversely, this behaviour is also seen in how some mobile applications and startups copy features from each other, claiming to "improve on" but only to end up overlapping and becoming clones of each other. I'm not fully adverse to the lean startup movement, but that's another discussion for another post.

3. The phenomena of experience and creativity

Perhaps the larger discussion that needs to be had is whether we found ourselves fearing for our professions because of the way we have come to think about our work. We have come to see ourselves as mere doers, the craftsman in a production line. The problem with this view is that, very often, a machine-manufactured item is far superior to a hand-made one (e.g. woven cloth, precision cutting).

The concept of an automated design process is not a new one. Camera lens makers have been using ray tracing software to automate the design process of camera lenses for decades. The lens designers input the desired optic specifications, intentions, even price points, and the software produces optimised designs. Yet, the best lens designers (like Leica) have come to realise that there are non-quantifiable and more human elements that computerised automation cannot satisfy. Elements such as experience and human creativity.

It is perhaps in these elements of experience and creativity, that highlight the real value of a designer. It is also likely that these are the very same elements that will be the most complex to teach a machine, considering we have difficulty teaching it ourselves—and can it even be taught? As Sir Alan Watts puts it in the contradiction of teaching creativity, "if we found a method whereby we could teach creativity, and everybody could explain just how it was done, it would no longer be of interest".

4. Parting thoughts

I don't think I gave Yong a straight answer, we eventually ended up chatting about something else, but as designers, perhaps we should invest time in improving our process, making our own processes and workflows more efficient. We should acknowledge that every practitioner is different and should spend time to tailor an internal process that addresses our own needs best. I find this is of greater interest and importance than trying to invent a copy of a present (and perhaps flawed) version of yourself.

Read next

Interview with HAYSTAKT →