Skip to main content

I recently heard someone liken artificial intelligence to the mystifying and transformative power of electricity during the 19th century. One can only assume that electricity was shocking and even scary to people of that time period; just as AI may be to us (just look at Elon Musk, who thinks AI is a bigger threat than climate change). As such, economists have predicted that automation of the service sector using AI will be 10 times as large as manufacturing jobs lost to automation. Most experts believe that human beings will still be needed for jobs that require high-order critical and creative skills… like graphic design. Or will we? There are more and more sites offering “generic algorithm” logo design. Can they replace legitimate designers? There’s only one way to find out, and that’s to get my own logo commissioned.

AI and Machine Learning

Before I start designing a logo, I want to give you a 30-second computer science lecture. Thanks to the likes of Google, Amazon, and Facebook, the terms “AI” and “machine learning” are now part of dinner table discussion. The issue is they are often used interchangeably..

Time for an example. When you are searching on your Google, and make a typo, it prompts you with: “did you mean..?” This is the result of machine learning algorithms in Google Search. Say you are searching for “cookies”, but accidently type “cookues”, you will go back and change it. Google’s algorithm makes note that you corrected yourself, and searched for something again, so quickly. It keeps this in mind for future users who make a similar typo. Google Search learns to correct it for you.

Artificial intelligence is the science of getting computers to act without explicitly being programmed to do so (basically, thinking for themselves). AI is the science and machine learning is the algorithms that make the machines smarter, which is a part of AI. They aren’t the same, but you can’t have one without the other… Confusing? I know… (Do you blame me for switching my major?)

AI and Logo Design

There are a number of services that are available for people who want to make a quick and dirty brand (Mark Maker, Brandmark, Logojoy). I will be using Brandmark. Brandmark proudly states that it, “uses deep learning tools to generate logos composed of an icon, typography and color scheme. The specific technologies used are conv nets, word embeddings and GAN”. Basically, Brandmark attempts to find unique icons, and pair them with a typeface using neural embedding. This process analyzes lines, and fills to find visually similar icons. It also creates a vector that encodes the visual information of a font (curves, serif/sans serif, etc). It then uses a combination of this data to match up fonts and icons with similar visual features using their neural embedding, which generally produces more cohesive logos. So, can Brandmark design a half-decent Muse logo? Let’s see.

The Process

Brandmark Company Setup   Brandmark Colour Choice

The first step of your logo journey is to choose logos you like from a set of over 20 logos. These logos will be used as “inspiration”. Then, you enter your company logo and “tell Brandmark about yourself”. This is basically the option to include a slogan. I went with “authentic engagement”. Next up, you need to enter at least three keywords about your business (“marketing”, “advertising”, “web design”, for example). Finally, you choose your colour palette. Obviously, I went with Muse orange. Brandmark does its magic, and generates icon and type combinations for you to choose. So are you ready to see the big Muse redesign?

The Results

Brandmark AI Results 1Brandmark AI Results 2Brandmark AI Results 3

These are the generated results, but you have the ability to further edit the logo you select. This includes colour, typefaces, switching icons, etc. My complaint here was that the icons all appeared to be very generic (and does a logo need to have an icon?). Brandmark also compiles “previews” of your brand. These are mockups of business cards, posters, etc., with your chosen logo. Finally, you can checkout, paying $55 for high-res .eps, .png, and even branding guidelines, or $75 for a responsive website template. So what are your thoughts?

Brandmark Business Card   Brandmark Mobile Website

I don’t think Muse will be choosing one of these logos anytime soon. Not that the program isn’t a nifty use of AI, but visually, the results are rather generic. Who do I think this software is for? Well $55 for a half-baked brand identity isn’t terrible, and I would recommend Brandmark for someone who is working on a personal side-project, or a startup just getting off the ground.

Nevertheless, if you’re serious about building a brand identity around your actual business and customer, you can’t replace the human with the robot. Brandmark can’t help you unlock potential with your brand, connect with potential consumers, help you understand new markets or engage with your audience. One can argue that all graphic design is derivative in some way. However advanced the library of fonts and shapes is, Brandmark doesn’t create something authentic; it simply adds to the graphic noise that we experience everyday. At Muse, we try to recognize this, and design authentic strategies around why you do something, not what it is you do.

Conclusion

Ultimately logo design is a very subjective field where attention to detail and communicating what your business is about, is key. I may be ripping these “AI” design programs now, but you know what they say: “any innovation that sucks today might put you out of business 10 years from now.” I don’t think it will replace legitimate designers in the near future, but this type of software opens the door to democratize the design process, making it accessible to more. Design doesn’t need to be too expensive, or overly complicated, but to me, it does need heart. It looks like the robots are missing that ingredient… at least for now.