AI Goes to School

By Aria Novosedlik

AI is everywhere these days. Annoyingly so. Like climate change, its ubiquity as a topic of conversation has become exhausting. Yet, like climate change, it must be discussed – thoroughly – by everyone, not just the tech bros at the top. 

If you need proof, look at Meta, which just last week appointed a group of outside advisors 'to provide guidance on its artificial intelligence strategy’. The group is comprised of four white men. Was this an oversight? Another case of cultural blindness? That’s precisely why we need to start encouraging discourse among all factions of people, particularly young people and students, who may literally be competing with bots to get a job. 

To get a better understanding of when and how these conversations are being held in design school settings, we spoke to Dominic Ayre, who teaches graduate level design students at George Brown College, and two of his thesis students, Ben Luu and Angelique Flores. Both Ben and Angelique focused their theses on Ai, though through very different lenses.

We asked Dominic how honest students are being with regards to their use of generative AI platforms Midjourney and ChatGPT, and whether said use was valid. All three agree that the use of ChatGPT is of greater concern than Midjourney, especially with the recent release of ChatGPT4o (see below). But as Dominic explained, ChatGPT also has great potential as a learning tool. For instance, there are many international students accepted into post-secondary programs despite their lack of English competency. They are actively pursued by post-secondary institutions due to the higher tuition they pay to be here. This is not to take away from their talent, but who can ignore the absurdity of making them sit through three-hour design history lectures in a language they don’t understand? AI can help with that.

Generative AI as an agent of inclusion

As Dominic notes, this is about inclusion, and AI can already remedy many translational issues, so the language barrier is evaporating. We’re even watching things like Google translate catch up and eliminate translator jobs within institutions. As mentioned above, few days ago Open AI launched ChatGPT 4o, the first multimodal AI. The ‘o’ stands for ‘omni’, a reference to its ability to understand and generate more than one type of data. For example it can process images and understand audio natively. But it’s going to need some serious DEI training if it can be used by all students. It’s imperative to ensure that not only is everyone involved in this process, but that everyone’s needs are considered in this process.

Results of a Washington Post prompt asking for images of ‘people who received social services’ An example of built-in bias if ever there was one. How would BIPOC people view this? 

Too many stories about inherent bias within AI have come to light recently. A few months ago, when The Washington Post prompted AI to generate images of ‘people receiving social services’, all results (above) were of non-white people. When prompted to produce images of a ‘productive person’ (below) middle-age white males in suits and ties were vomited up. 

In a glaring example of tokenism, the Washington Post’s recent experiment with AI reveals our stereotypical view of productivity: apparently it’s very messy and almost completely white.

To say that’s cause for concern is an understatement. Unfortunately, as Dominic agrees, the adoption of DEI as common practice has not happened in parallel with AI; AI’s learning rate far outpaces our ability to achieve social change. It seems nobody has an answer to this problem yet. 

As for the matter of  transparency and ChatGPT, Dominic mentions that to his knowledge, only a small number of students used it to produce text without disclosing to him. A few others were up front about their use of it and made it clear that they were experimenting with it as part of their research (e.g., Ben and Angelique). His stance is that within academia, we should develop policies and be realistic with the fact that students will take a lifeline when offered. With that in mind, we need to be more considerate of AI’s specific applications within the context of education. Dominic points to Toronto Metropolitan University, which has put together a manual for all use cases of AI at their institution.

Like most designers Dominic refers to Midjourney as plastic-y, low-hanging fruit, at least for now. But others have been preoccupied with how quickly it will evolve, fearful that their passion for design will be rendered irrelevant by bots. Dominic notes the fact that, as with sugar vs artificial sweetener, it's easy to tell the difference between a Midjourney hand and a handmade hand: the Midjourney hand is the one with seven fingers. But AI is still in diapers. For example I am a casual life drawing artist and I still haven’t mastered hands and feet. But it’s only a matter of time until AI does.

For his graduate thesis at George Brown College, Ben Luu probes the type design capabilities of AI with a font named ‘People (as seen by computers)

The speculative future of designing with AI

Ben Luu didn’t start off in design. He has a computer science degree from UofT, yet when he pursued design at George Brown and chose AI as the theme for his thesis, he decided to shift focus from its technical aspects and instead examined AI through a humanist lens. 

Cover and spread from Ben Luu’s ‘Computers Are Like People

He wanted to see if there was a way to collaborate with AI to design a typeface. ‘AI is so difficult to use and because it doesn't have enough autonomy, we can’t be reassured that design is still going to maintain its relevance in the hands of people.’ There is a palpable urgency to his tone—very understandably so. He isn’t thinking about whether Midjourney can draw hands in a few years. ‘I'm much more interested in where we see AI developing in the next 10, 20, 30 years’. 

Spread from ‘Computers Are Like People’: is this the mobile interface of the future?

His concerns are far more abstract and philosophical: the nature of authorship, the nature of art, the nature of legislation, the political side of technology and controlling bodies of power. "These are much larger questions that, in the context of how academia approaches AI, will have to be addressed not through policies, but through mandatory courses for degrees that aren’t even specifically related to AI. 

These pages reveal a moment of optimism in Ben’s essay about the future of ‘handmade’ design in a world of bots

"Like design history, the use of AI will need to be taught as a design sub-discipline in the future. If we’re going to teach students about design’s past, why wouldn’t we teach them about the potential of design’s future—one which will inevitably involve AI?" 

Type as foresight and speculative design: Is this what ‘far ahead’ looks like? 

That’s at the heart of Ben’s argument. We aren’t looking far enough ahead, and even when we do, we get caught up with the big sexy problems instead of thinking about the more realistic implications it poses and the surrounding social impact. When asked whether he feels more positive or negative about AI, Ben bluntly states ‘negative’, though he agrees that education is one realm in which it could have a net positive impact if implemented with more care.

When does innovation end? From Angelique Flores’ grad thesis ‘Amalgam’

Is curation the future of design?

Angelique Flores’ thesis, ‘Amalgam’, has plenty to say about her personal views on AI. Her intent was to see just how well AI can parse and reproduce. “It was a 12 week experiment of training two different AI models to see if they could capture two distinct design styles. I trained it with 40 uses of Swiss International style, as well as 40 instances of Polish poster style. I used that and added prompts to see how well it could generate new imagery — to see if it feels original, if it’s interesting, and if it stands out from what you currently see with AI training models.” 

Are we ever going to be able to catch up with the speed and power of our own technologies?

Like hands, typography has yet to be mastered by AI. Because graphic design relies on type and abstraction, we needn’t worry about job loss just yet. Nor do profs need to be concerned with the possibility of work submitted from students that haven’t taken any part in the process of creating it. 

It looks like a human, talks like a human, and writes like a human: will we be needed to design in the future?

Angelique, like most of us, understands that even if half our work is cut out by AI, there remains the need for designers to curate. The need to teach fundamentals like typesetting will persist in the short term, and truly passionate design students will always be drawn to the tactile aspect of designing.

The beauty of hesitation: a sign that there is thinking going on here

Dominic — along with Ben and Angelique — all noted that among profs, there really hasn’t been a consensus as to how AI should be accepted and to what degree. Some are quite against its use, while others are aware that a policy of honesty is far more productive. Regardless, between Dominic’s desire to create a hard manual that elucidates AI’s use cases, Ben’s hope for education systems to gaze into the future with a humanistic eye, and Angelique’s uncertainty over how successfully AI can really replicate what the human head, heart and hand are capable of, the only thing that can be said with any certainty is that the 'design with AI' genie is out of the bottle, and there’s no putting it back. 

Aria Novosedlik is a designer, artist, researcher and writer based in Toronto