AI is Taking Over Our… Typefaces? Oh, That’s Not so Bad.
I have been designing Typefaces for about six years. I have been using AI to help me design typefaces for about 6 weeks and I could not be happier. Hot Take, but don’t be surprised if you are thinking the same way by the end of this article.
When it comes to artificial intelligence and design everyone has their opinions and bias so imma just get things out of the way right quick with no illusions and tell you my bias and where they come from. I got 3 basic opinions on AI
- Use AI if “average” is your goal.
- It is a great way to “Put Clothes On” your craft.
- It will take the jobs of those unwilling to integrate it.
After I quickly address each one these biases and where they come from I will then talk about how AI and its rapid growth are affecting the world of typography and how I predict things will play out in the future.
To average to go under to average to go over.
Understanding expels fear. Many who fear the capabilities of AI do not understand its inherent limitations. The nature of AI (in an extremely simplified way) is this: Filter, Collect, Synthesize remix. The user prompts the AI, essentially applying a filter to what AI should collect. AI then searches the internet (or one part of the internet depending on how it is filtered) then collects information. With that information now collected AI synthesizes what pieces of data are valuable in that collection and groups them together into basic averages. Then remixes those averages into a new iteration of that data.
So what does this mean on a practical level? AI is only as good as the filters applied, and the information available. Let’s apply this in the world of typography. Let me introduce you to “Joe Business Owner”. Joe, instead of buying a typeface or using a free typeface for his business for some reason, decides to use AI to make a typeface. Because Joe Business Owner is a business owner and not a typeface designer he does not know the terminology he might use to apply a filter that will give him a collection of information that is nuanced enough to only gather high quality. He simply tells AI to “make me a clean sans serif typeface” so AI searches the internet for all “clean sans serif typefaces” then synthesises and remixes them into something new and…
Cliff hanger as we zoom out for a second and look at a quick history of digital type design. (Don’t Worry we will get back to Joe in a bit) About 6 years ago the bar of entry to making a typeface drastically lowered. It seems like ever year there was a new illustrator plugin or some app that allowed you to “make your own font” any Shmoe with an ipad or an adobe subscription could now make a font. (I too was one of those Shmoes)
But with the ease of making a typeface at an all time high The easy of access to high quality typeface design education that was about the same as it had always been… very low. The result was what you might expect. A lot of really really poor quality typefaces were made. So now when AI looks for typefaces on the internet it finds a lot of really really poor quality typefaces. It collects those because it has no way of discerning the design quality of a typeface. Then it synthesizes and remix them into an average of all the typefaces in its collection. And in the case of typefaces the average of all typefaces that exist on the internet is very very far below the quality of work a well practiced typeface designer can produce.
This is true for most any creative field. Right before AI rolled up to the scene the ease of digital creation was at an all time high resulting in the barrier to entry becoming almost non-existent meaning the quantity of low quality work that exists on the internet is at an all time high.
So now, when a person is not learned enough in a particular field of design to apply a filter to an AI that can reliably collect only high quality work, that person ends up with collection results that will automatically include mostly low quality work. Therefore the remix of the average of that work will be “average” at best. Because of the low barrier of entry pre AI average work generally means low quality work.
Now back to Joe Business Owner. His AI is done doing its thing and he now has his master piece “JBO Sans” JBO sans is entirely average and Joe loves it. He uses it for his brand and every designer hates him. But Joe doesn't care. Because Joe sells expensive pest control to doctors and lawyers who live in big houses. And doctors and lawyers who live in big houses do not give a flying flip about high quality typefaces and just want no bugs in their house. So Joe makes a million dollars and loves AI.
Now you might look at this through a cynical lens. But I do not. If AI can do something better than you, and if the quality of that thing being average is acceptable to you: by all means, use AI to do that thing.
This is a good gut check for when you should and should not be using AI. Ask these two questions.
Can AI do this thing better than me?
Does the quality of this thing need to be only average?
If the answer is YES and YES: Use AI.
If the answer is YES and NO: Hire an expert to do that thing.
If the answer is NO and YES: Do it yourself
If the Answer is NO and NO you are somewhat of an expert in your field and you should be doing that thing for other people who value that thing.
This naturally leads to my second bias
AI is a good way to “Put Clothes On Your Craft”
The best way to explain this ridiculous metaphor is with a personal example. I Love drawing letters and I am confident in saying my ability to do so is above average. I also sell the typefaces I make. To me it is important that the drawings of the letters in the typefaces I sell are above average So I never use AI to draw letters. It just doesn’t make sense. However, There are certain aspects of making and selling typefaces that I am okay with being just average.
For example, creating small caps is a very tedious process. Once the regular capital letters are drawn they need to be copied, renamed, scaled, the sidebearings need to be readjust, the Kerning groups need to be renamed the kerning pairs need to be adjusted and blah blah blah blah I lost you a while ago. All this to say I used AI to create a script to do all of that for me. I am not very good at coding. AI is for sure better than me at coding. And when it comes to copying and renaming and scaling and all that there is no issue of average. It is binary. Yes or no. Did the character copy? Was it renamed correctly? If so, all good. I then go through and fine tune the drawings of each of those newly created characters.
This script took me about 2 days to refine working with Chat GPT. Before the script this process with all these steps took me about two days. Now with the script the process takes me about 30 seconds.
I also use AI to help me code my website in a similar way. I am not expert at web design some tasks I can do better than AI like the designing layouts, Others I cannot Like the implementation of the design and and coding of that website. There are some aspects of my website like the poster maker that I could not do myself but needed to be of a very high quality so I hired an expert (Cotton Studio) to make that thing.
I use AI to help me work on SEO and marketing strategies for my business. Because I hate doing it and am okay with an average strategy to get things started. This gives me time to focus on the things that I do better than I am and things that need to be done at a better than average quality.
The coders, and the SEO people and the marketing strategists who deliver average results probably hate me for not hiring them the same way they hate AI for taking their job. But like Joe Business Owner I am fine with that because in the same way rich people with bugs in their houses don’t care about fonts people who value well designed typefaces don’t care about SEO strategies.
At the end of the day I draw the letters. That is my craft. All the other stuff is just clothes my craft puts on. Clothes don’t make a person who they are. They just make them presentable in certain settings to certain groups of people. They can be changed, the style can shift over time, the quality of the clothes can vary. But the clothes. In the end do not define the person. The same way that AI should not define the craft.
This is why I believe that businesses built using “AI Hacks” as a foundation are shallow and ultimately will not be as successful as businesses with human craft as a foundation.
The industrial revolution shifted culture in a similar yet opposite way. Hyper specialization and mechanical repetitive production became valuable. Each person on an assembly line was an integral part of the equation. However each person on the assembly line could just as easily be replaced by another. However the person who designed the assembly line was never worried about an assembly line worker taking his job. Both the assembly line worker and the designer of the assembly line “made cars” but they did it in two entirely different capacities.
The AI revolution differs from the Industrial Revolution in one key aspect. AI gives the line worker the ability (if they are willing) to become the person who designs the assembly line.
The “Means of Production” are now available to anyone with an internet connection.
That leads in nicely to my final bias
AI will only take the jobs of those unwilling to integrate it.
Jumping back on the assembly line metaphor. Lets says a line worker living through the industrial revolution loses his job. That is tough, it’s hard when anyone loses a job at any capacity. But let’s say right after he gets fired he is approached by a stranger who says “hey friend, I have built a successful company that manufactures automobiles and now I want to teach you how to do the same I will give you the knowledge to do so and the money you need to build the factory all you have to do is come up with the look of the the cars.
How upset do you think that line worker will be about losing his job now? It’s a totally irrelevant issue. He can take up the offer of the stranger and live out his days designing and manufacturing cars.
That is the same scenario we are all facing with the AI revolution. Yes it may take our jobs as production workers but at the same time it puts into our hands the “Means of Production"
Back in the day if I wanted to run a type foundry I would have had to own factories and buy lead and hire workers to produce lead type, and figure out supply chains to deliver those boxes of lead type. In 2020 I bought glyphs mini for $50, watched a bunch of videos on the internet and now I run a type foundry. AI makes the bar of entry into many industries extremely low. The effective bar of entry into owning a business is an internet connection. So yes AI will probably take the job you once had or maybe currently have. But in the same instant it offers unimaginable opportunities for those brave enough to integrate it skillfully.
Okay Earostotal… So What?
Now that we got the philosophical part out of the way let’s take just a moment to think about how these ideas play out in the world of typography. First, it is important to recognize that digital type has always served two markets; consumers and businesses. These two markets buy typefaces to accomplish the same task. To write words. But, the “why behind the buy” is extremely different.
Consumers or “hobbyists” just want a little font to make a little thing. Pre AI, your aunt who is making shirts for a 70’s themed family reunion used to have to go to creative market and buy a $16 font, install that font on her computer, then typeset “The Bickmores” on a shirt. Your aunt is not a designer so this process is not an easy one. Now, POst AI, all she has to do is type “make me a groovy 70’s style shirt design that says ‘The Bickmores’” AI does this faster, better, and for less money than she could. And she doesn't really care what the shirt looks like as long as it’s groovy so average is acceptable.
AI will drastically diminish the demand for display typefaces sold at low costs to hobbyists. If you are a type designer in this lane pivot now.
Businesses are a whole different story. Businesses manage brands and brands need to use typeface not just for one project but over and over hundreds and thousands of times across years and years on a variety of different applications. Businesses buy typefaces as much for their utilitarian value as for their aesthetics value. Large companies know that typefaces are more than just digital pictures of letters. Typefaces are software that need to be integrated into every part of a company's infrastructure. Typefaces of this ilk need special attention given to not only the drawings of the letters but the coding of the typeface software.
Besides the practical utility of a typeface there are also legal restrictions on how large international companies need to integrate typefaces on things like their websites and apps.
Imagine the amount of nuance it would take to create a prompt or series of prompts that could produce a typeface of this caliber.
Trying to prompt an AI to filter through all the bad typefaces and create a collection of typeface to reference–all of which include the technical and legal nuances required by typefaces suitable for large brands is impossible. This is why Most type designers who achieve this level of mastery have legal contracts backing their typefaces that include a section on how their typefaces may not be used to train AI models. Any free typefaces of this caliber such as some typefaces found on google fonts run into legal problems as they exist in one degree or another in the public domain and cannot be owned and insured by the company using them.
In order for a prompt of this caliber to become a possibility enough typefaces would have to exist without the legal clause preventing AI from using them as to train models. Or Laws would need to be passed that make all of those clauses void. Then a prompt of this magnitude would be possible. In which case in order to actually create this mega prompt one would need to dedicate years to the study of typeface design to be able to pull this off. Expecting anything less would be as silly as prompting chat GPT to “make me the next big social media platform” and expecting a fully functioning spiritual successor to instagram, facebook, or tiktok. It’s just not going to happen. It takes teams of dedicated craftsman who consistently produce above average work years and years to make something like that happen. Therefore It is much easier for large companies to spend the money to buy typefaces that fulfill all their needs.
If you are a typeface designer focused on creating typeface for this type of client AI will help you rather than replace you.
It goes back to the two key questions.
Can AI do this thing better than me?
Does the quality of this thing need to be only average?
Your aunt making shirts answers YES and YES so she uses AI
Large Companies answer YES and NO. So they hire experts.
The number of typeface designers creating typefaces that are appropriate for use by large companies in a branding capacity is very small. And with Monotype changing to a mainly subscription based licensing system more and more brands will look for other solutions. Brands who are fine with average will use AI and brands who require above average quality will look to independent type founders who can serve their needs at a more affordable cost with no monthly subscription.