Post

Having students read @n-w-f.bsky.social’s 2009 writing on the Eliza effect and then comparing to the contemporary “GPT effect” including exercises and assignments trying to poke at a LLM chatbot to suss out limits of its abilities, tendencies, training data, etc.

Now each other panelists presents an exercise in this area, starting with textile-based play designer & educator @codingcrafter.bsky.social who starts off by summarizing higher ed’s polycrisis, and how everyone (admins, faculty, students) urgently needs more thorough, critical AI literacy

Oooh it turns out that game design educators are the best people to teach students about this topic, for all these reasons. I mean, makes sense to me (incredibly biased tone of voice). Gillian taught an AI, Design & Society class in a first year “great problems” seminar at WPI

The project was a reflexive/meta idea! Design a game to be played by every other first-year students, considering a future version of their school and how AI should or shouldn’t play a role. A couple games described are about tradeoffs (AI destroys academic workload faster but is less accurate)

…or e.g. a deck-builder simulating loss of critical thinking ability due to AI. Students were also involved in co-creating AI policies for class, to avoid prohibition dynamics and foster conversation. Mentorship from experienced students, playtesting and help desks (“benefits to going to a human!”)

Next up @recardona.bsky.social starts off by talking about Joan Morgan’s “When Chickenheads Come Home to Roost” and the coining of Hip-Hop Feminisms out of a need to “fuck with the greys” (nuanced ambiguities, resist the need to solve problems, trust in lived experience)

Central problem — the AI Creative gap. Ability to produce what resembles creative material rapidly outstrips taste, the ability to evaluate, understand, appreciate and discern creative material. Rogelio pointing out that a lot of educational assessment relies only on quantifying output, not process

So, one way to go would be to shift every wholly into process education — lots of cycles of feedback and working through problems together, qualitative evaluation, etc instead of evaluating output. This is expensive! Shown, a student party game dealing in the messiness of “output”

Next up @cheesetalk.bsky.social who couldn’t enter the US in person; showing her 1001 Nights experimental game which uses a LLM King Shahyrar to evaluate the Sheherazade character’s stories. Talking about how vibe coding has affected her collaborative process with coders & artists

They have non-technical designers produce a vibe-coded prototype along with a design doc to communicate intent, then handing off to a Unity programmer. Unity programmer puts the design doc in context to guide code too

“Beware the beautiful chicken” — vibe coded demos look deceptively finished, when they may have no valuable gameplay.

Finally it’s Universal Paperclips dev and guy who used to do my job & says he still has opinions about teaching, @flantz.bsky.social says: 1. It’s a disturbing time but WE’RE the lucky ones; our field is already about intersections of feelings and code, taste and software, trying to find meaning.

“Even if AI is making you miserable and depressed — to be an artist who’s present in a historical moment where things are happening and something is at stake, that’s a kind of luck.”

It’s also our job to think about this stuff; game designers and all the artists who work with code are the R&D department of computation; we explore the limits of how it can mean things. So we have to take the opportunity to engage with these issues, especially if society has subsidized…

…an institution you work at, like higher education. Where the reason we have jobs is in part to figure this shit out. And it turns out to be MUCH harder than expected. In 2020, Frank thought AI would open a window into new kinds of games & experiences. Newly available forms of computation…

…that could be mind-blowing, psychedelic, create new surfaces for play, new possibilities. Now, years later he has done a whole lot of experimentation in this area, started a studio, etc and has NOT found this promise fulfilled; there is no powerful, transformative opportunity that he once expected.

On one hand, maybe this is just him, and there are some cool weird little experiments, but overall he looks around him and does not see anyone accomplishing new forms of gameplay with AI that have the necessary, recognizable spark that would accompany genuinely amazing new experiences

Why not? First, the business models are messed up for “AI native games”; “pay as you go” token costs for inference create bad incentives (“designing around cost”) and unaffordable virality for new interesting ideas. Also, model change is a rug-pull for iterative development.

Second, lots of players have decided that AI is low-quality cost-cutting (not to mention, for some players who care, possibly labor replacement) and so there’s a widespread controversy attached to use of AI in games (no explanation necessary, c.f. Expedition 33 etc)

But perhaps most importantly, a theory about THE VERY NATURE OF FUN — this new form of computation is diaphanous and ephemeral (jiggly?) and do not really work with what with think of as “game mechanics.” (How I might put this: LLMs are bad at producing consistently structured output?)

The clockwork part of the system is simple and boring and maybe awkward on its own. But the experience which is produced by these “simple machines” is irreducible and mysterious, very human, affective, etc. You can’t stuff THAT kind of thing INSIDE the machine to make it work, basically