Mixfont Update - seeking a designer to help complete an AI-generated font
Comments
-
I am an employee of Google USA, in the font development lead role on the Google Fonts team, and this is personal opinion not an official company statementMan, this is depressing. I can't imagine what it must be like to moderate these submissions.I think this is a typical arms race situation, where the AI-powered creation/submissions are to be met with AI-powered moderation/reviews, just as how upon receiving a 12 page email written by an LLM based on a one paragraph prompt, email clients then produce a 1 paragraph summary...
0 -
Hilarious.1
-
That sounds like turbo-accelerated enshittification. Why not simply ban AI-generated submissions? The quality is abysmal, people who do the submissions have no idea how to take on feedback, since they have no idea about what they produced. This is futile.
9 -
I hope it doesn't come to this. A lot of the submissions meet the technical requirements (i.e. all the required glyphs and Fontbakery ticks) but are poorly drawn and assembled, especially diacritic handling and any glyphs outside of the Latin Core glyphset. See Exhibit A from a recent submission.I think this is a typical arms race situation, where the AI-powered creation/submissions are to be met with AI-powered moderation/reviews, just as how upon receiving a 12 page email written by an LLM based on a one paragraph prompt, email clients then produce a 1 paragraph summary...
But perhaps Google has access to this new 'super' intelligence (lol) that is capable of pointing out the mistakes of plain old artificial intelligence.
0 -
That has certainly been true so far.SCarewe said:The quality is abysmal, people who do the submissions have no idea how to take on feedback, since they have no idea about what they produced. This is futile.
But it is entirely plausible that quality will get better in the near future. So “not looking at them” seems like a bad choice if one wants to know how things are going.
I made one incredibly wrong background assumption when I prognosticated about this a year ago, at last year’s Fonts & AI conference: I figured that the folks investing in AI generated fonts would have somebody involved who had expertise in type design and font quality. For everything that has rolled out in the western world so far… that seems to have been false. What a huge handicap that is for the developers! (I expect Chinese and Japanese may be further along... may learn more in the next couple of days.)
I mean, yes we are seeing a bunch of AI-generated western fonts. It took a little longer than I estimated (I said “within six months”). But I am blown away… by the lack of quality.
That doesn’t mean things won’t improve. I fully expect that. TBD how much so, and how fast.1 -
We've come full circle back to the original post 😅.I figured that the folks investing in AI generated fonts would have somebody involved who had expertise in type design and font quality.0 -
Not at all. Eric is asking somebody to fix up a single font, so he can get it accepted as “the first AI font in Google Fonts.” Nothing he has written about it suggests to me that this is an effort to improve his production chain. Genuine feedback on that has… not gotten any response.0
-
Thomas Phinney said:
That has certainly been true so far.SCarewe said:The quality is abysmal, people who do the submissions have no idea how to take on feedback, since they have no idea about what they produced. This is futile.
But it is entirely plausible that quality will get better in the near future. So “not looking at them” seems like a bad choice if one wants to know how things are going.I disagree with this. it's clear they'll get better, possibly indistinguishable from human-made work, but that skips the more salient half of sebastian's point RE: people not knowing what they're producing. fonts are more than a set of vector drawings and metadata encapsulated in an OTF file, the end result isn't the full story, and AI-generated marketing won't fill that gap either. people dedicate their lives to this craft, and if I were a distributor, especially one with as much brand equity to lose as Google Fonts, I should want to be a gatekeeper, and to not become known as a cesspool by evaluating the work of those who do not have a craft involved in their process. an AI-native Cesspool Fonts (or Mixfont, or whatever it ends up being called) will move faster than Google anyway...1 -
Simply, simply.SCarewe said:Why not simply ban AI-generated submissions?
Should fonts with kerning from kern-on also be banned?
Closer to home, how about fonts with a generated italic?
I don't mean to be so facetious, but over the past 20 years of participation in this community, I've seen ever increasing productivity through ever better automation tools; and I've attended plenty of conference talks by people lamenting the arrival of DTP and Fontographer, or even Unicode, starting 26 years ago when I began formal schooling in graphic design.
0 -
Also, how do you ban AI generated submissions? Having a big sign saying "Google Fonts doesn't accept AI-generated submissions" will not stop actually people from making the submissions - there is a submission form already which asks people to commit to disclosing AI use, and if you browse the issue tracker you will discover that many people just tick all the boxes in the form without reading them. (Including the one which asks them where the actual font sources are...) And so even if you ban AI-generated submissions they will come, and they will need to be evaluated before they can be dismissed, and that will take some kind of effort from somebody somewhere. The slopocalypse is real - it's Brandolini's Law applied to fonts. At some stage, you gotta fight fire with fire.There also seems to be a real mix. "AI-generated submissions" is not a binary. There is clearly slop, people who have had an idea for a font and got the machine to do all the work, and I find it fascinating to examine the different ways that the robots are trying to turn the idea into a font (this is a particularly interesting example of a bot really trying hard, but ultimately not really succeeding); there seem to be quite a lot of people who are using AI to help with the engineering and release readiness, which I guess is OK, but man, reading the word salad gets tiresome, and just following the GF Guide gets you a lot further without the friction; and there are a number of people who are using AI to help communicate because their first language isn't English, and actually I'm pretty comfortable with that on the one hand because helping people to communicate is good and lack of English ability should not be a blocker for participation, but there's a real downside in that you don't get a sense of who they are, what they're trying to do, and what their passion is - all their communication is moderated through the usual "helpful" codebot personality. (Unfortunately "Two things: a suggestion, and a request" does not inspire me to help with either.) There also seem to be a few people who want to make new fonts that are slightly modified versions of old fonts, sometimes with bug fixes or extended language support, and I just think "why not fix the existing font?"
Weird space race indeed.
(Disclosure: I'm a contractor, not an employee, of GF, so this is purely my own opinion. I'm not responsible for moderating the submissions queue, but I do browse the repo and note the odd technical issue.)
8 -
These are determinisitic algorithms, as you know. They require a knowledgable designer's eyes and input to manipulate. These tools aim to accelerate the manual processes that are formalisable as algorithms (according to the tool developers, who bake their own, personal approach into their tool) and are designed as time savers, not as replacements for a creative process. They are not trained on design output.Dave Crossland said:
Should fonts with kerning from kern-on also be banned?
Closer to home, how about fonts with a generated italic?
(...) increasing productivity through ever better automation tools (...)
People who use Kern On or Italify know (more or less) what they're doing and will be able to take on feedback. I have seen plenty of bad kerning with Kern On by people too lazy to learn the tool properly (or who have too little experience kerning manually). Italify also delivers bad results if used improperly, of course. And yes, of course fonts should be rejected that are improperly delivered using these tools, or any other tools, for that matter.
The difference is simply that AI allows people to generate fonts with zero prior knowledge. The bar is inexistent, whereas other type design tools lower the bar in specific areas. I think this comparison makes no sense.11 -
this is really not the GOTCHA you hoped for. Automation in the process (eg Kernon) is one thing, using essentially gen AI trained on industry people is wildly not the same.Dave Crossland said:
Simply, simply.SCarewe said:Why not simply ban AI-generated submissions?
Should fonts with kerning from kern-on also be banned?
Closer to home, how about fonts with a generated italic?
I don't mean to be so facetious, but over the past 20 years of participation in this community, I've seen ever increasing productivity through ever better automation tools; and I've attended plenty of conference talks by people lamenting the arrival of DTP and Fontographer, or even Unicode, starting 26 years ago when I began formal schooling in graphic design.4 -
It took until yesterday for me to realize that if basic search results had not by now become a dog's breakfast of pop-ups, inline ads, paywalls, and other long-established enshittification, there would be no demand for things like chatbots to comb through it for us. As it stands, I ask Grok a question every other day and tend to get very useful answers.1
Categories
- All Categories
- 47 Introductions
- 4K Typeface Design
- 501 Type Design Critiques
- 593 Type Design Software
- 1.1K Type Design Technique & Theory
- 675 Type Business
- 910 Font Technology
- 29 Punchcutting
- 544 Typography
- 129 Type Education
- 333 Type History
- 82 Type Resources
- 114 Lettering and Calligraphy
- 33 Lettering Critiques
- 81 Lettering Technique & Theory
- 584 Announcements
- 101 Events
- 117 Job Postings
- 179 Type Releases
- 186 Miscellaneous News
- 270 About TypeDrawers
- 54 TypeDrawers Announcements
- 114 Suggestions and Bug Reports





