Matthew Guzdial
matthewguz.bsky.social
Matthew Guzdial
@matthewguz.bsky.social
Associate professor @ University of Alberta and Canada CIFAR AI Chair @ Alberta Machine Intelligence Institute.

Games, machine learning, and creativity | he/him
Thinking a lot these days about the parallels in underfit models in humans and ML. “All <group> are like X” is basically posterior collapse, and both arise out of struggling to deal with real variety in experience/data. Intellectually/computationally “lazy”. Has someone written about this?
December 29, 2025 at 4:39 AM
I really wish the author of this piece (not Bruno) had talked to someone about how world models actually work. They are wildly infeasible as a disruptive technology for game development unless everyone owns a TPU (not a GPU, a TPU).
Literally everything the principals involved here say is just word salad nonsense that doesn't even approach meaning enough to be a lie, but it's also worth noting that the game given up as an example of this tech has 500k downloads on the Android app store. www.ft.com/content/9b1b...
AI ‘world models’ promise to reshape $190bn video games industry
Google DeepMind and Fei-Fei Li’s World Labs target gaming with AI-generated 3D environments
www.ft.com
December 28, 2025 at 4:16 AM
Reposted by Matthew Guzdial
definitely the most flexible and magical level design tools I've ever used! @2darray.bsky.social is a 3d math genius
What was the level design process like for BIG HOPS and what tools did we use?

Question from Terri V on Bluesky
#indiegames #gamedev #bighops
December 27, 2025 at 5:24 AM
Reposted by Matthew Guzdial
I had five fascinating conversations for this RPS piece, gathering perspectives on genAI chatbots in games, as seen in Where Winds Meet. Thanks to @meghna.bsky.social, @pyrofoux.bsky.social, @maxkreminski.bsky.social, and @chrisgardiner.bsky.social www.rockpapershotgun.com/horrible-bor...
"Horrible", "boring", and "cheap": Experts pan new chatbot NPCs, but some leave room for optimism
We speak to a collection of developers and AI experts about the potential impact of LLM-powered NPCs in videogames.
www.rockpapershotgun.com
December 23, 2025 at 4:22 PM
Reposted by Matthew Guzdial
Who is college for? And what does college do? are two questions of increasing urgency because if those of us involved in these institutions don’t provide compelling answers then the tendency towards reducing it to glorified job training will destroy every meaningful part of the endeavor.
December 18, 2025 at 11:37 PM
Reposted by Matthew Guzdial
We are studying the sentiments of visual artists towards generative AI in the workplace and their impacts on creative careers. If you're an artist, please consider filling out this recruitment form for access to our survey!
cmu.ca1.qualtrics.com/jfe/form/SV_...
December 19, 2025 at 1:58 AM
What'd I do for my birthday? Well I just submitted a 2 million dollar grant*. Now to pack for India tomorrow before serving on an Australian PhD Committee Defence at 11pm tonight.

*(translator's note: for US audiences this is the equivalent spending power of a 20 million dollar grant)
December 16, 2025 at 6:53 PM
Reposted by Matthew Guzdial
I will unfortunately say this for the game awards: there is an increasing body of proof that simulshipping a big content update with your big ticket win yields pretty substantial sales for already-successful games
December 15, 2025 at 9:21 PM
When you write your dozenth unique grad school reference letter of the season they should let you access the Secret Reference Panel where I can just say "accept this person".
December 14, 2025 at 10:51 PM
Reposted by Matthew Guzdial
Since Polygon is leaving it out of the headline: Epic Games partnered with The Game Awards to put money into the pocket of a bigot who uses her wealth to further the genocide of trans people through lawsuits and legislation. Every 'free' game gives money directly to JKR, and they are complicit.
December 13, 2025 at 2:31 AM
Reposted by Matthew Guzdial
I know everyone is saying AI is doing this but I think it's the entire economy sucking shit that's doing this and they're just using AI as an excuse.
November 18, 2025 at 4:16 PM
Reposted by Matthew Guzdial
Hi friends! This weekend, I'm planning to share deets about a #ttrpg project I've been working on for over a year now <3

In advance, I truly apologize for what's about to happen, muahahaha ✨

❌🎲👁️
December 11, 2025 at 5:58 PM
Unless they‘re nationalizing Google, the US national security threat here is insane. Some Google server pinging with the locations of *all* of your armed forces? Scraping training data of deployment plans?

If you do want a chat assistant for the military just make your own internal LLM, it’s easy!
Pete Hegseth: "The future of American warfare is here, and it's spelled AI."

He says the military has launched a new platform that  "puts the world's most powerful frontier AI models, starting with Google Gemini, directly into the hands of every American warrior."
December 9, 2025 at 8:25 PM
Reposted by Matthew Guzdial
Love to see the Sedgwick County Zoo putting out a good explainer on why GenAI slop videos of animals are a problem & how to identify them! It’s succinct, easy to read, and doesn’t shame people for not knowing better. A+ useful link to share with family and friends.

scz.org/blog/the-rea...
The Reality of AI Animal Content – Sedgwick County Zoo
If you spend any time on social media, you’ve probably seen them – bears bouncing on trampolines, apes caught on doorbell cameras, or “rescued” wild animals acting in ways that seem too good to be…
scz.org
December 9, 2025 at 2:44 AM
Reposted by Matthew Guzdial
Out now: WHALE ROADS

The chronicle of a year-long, 30+ player West Marches campaign of Wolves Upon the Coast by @lukegearing.bsky.social

A 250-page "zine" filled with play summaries, original art, and Referee commentary.

Available now, and it's 100% completely free.

sulcata.itch.io/whale-roads
WHALE ROADS by sulcata | Brendan McLeod
A megazine about Wolves Upon the Coast
sulcata.itch.io
December 9, 2025 at 3:06 PM
I typically don’t post my low fodmap experiments on here but when @persephone.blue says it’s in their “top five cakes” how can I not share my custom low fodmap carrot cake?
December 9, 2025 at 5:02 AM
Reposted by Matthew Guzdial
This looks really interesting. Seems primarily targeted at cog-sci and/or linguistics research, with Portal 2 just the example problem-solving domain, but probably also of interest to game-studies and game-AI people.
A couple years (!) in the making: we’re releasing a new corpus of embodied, collaborative problem solving dialogues. We paid 36 people to play Portal 2’s co-op mode and collected their speech + game recordings.

Paper: arxiv.org/abs/2512.03381
Website: berkeley-nlp.github.io/portal-dialo...

1/n
December 6, 2025 at 4:04 AM
Reposted by Matthew Guzdial
Spotify is garbage on every count: Its treatment of artists, its ICE advertising, the CEO's investment in military AI, its leading role in the commodification and AI slopification of music, its terrible audio quality—you name it.

So I quit, and put together a complete guide to getting off Spotify:
How to quit Spotify
This Black Friday, here's a guide to finding the best Spotify alternative
www.bloodinthemachine.com
December 3, 2025 at 12:09 AM
Reposted by Matthew Guzdial
For this year's ProcJam, I made a co creative tool to generate harmonies using genetic algorithm. The user first records a melody, and then acts as a fitness function by rating the harmonies. It is a little finicky but I had fun making it!
spiderassassin.itch.io/genetic-harm...
#PROCJAM
Genetic Harmonies by spiderassassin
Play in your browser
spiderassassin.itch.io
December 3, 2025 at 11:14 PM
Reposted by Matthew Guzdial
OpenAI’s big idea is to teach the LLM to post-hoc explain how it solved a problem. This is an extension of chain of thought.

It looks very similar in nature to “rationale generation”, an explanation technique that has been around since 2018.

www.technologyreview.com/2025/12/03/1...
OpenAI has trained its LLM to confess to bad behavior
Large language models often lie and cheat. We can’t stop that—but we can make them own up.
www.technologyreview.com
December 4, 2025 at 2:12 PM
Reposted by Matthew Guzdial
Hey #YEG friends! Are you involved in indie TTRPGs? Do you love running and designing and playing #TTRPG games? We've got a meetup for local indie TTRPG folks at @pemetawe.bsky.social on December 27 at 1:00 PM :D Be there or be too cool for this XD
December 3, 2025 at 2:22 AM
Reposted by Matthew Guzdial
My former MS student Chris Cui (now PhD student with @rajammanabrolu.bsky.social)motivates Text Adventure Games as testbeds for reasoning. Provides a new benchmark suite of text games. Observes that Zork still kicks LLM’s butts despite training on walkthroughs arxiv.org/abs/2504.14128
TALES: Text Adventure Learning Environment Suite
Reasoning is an essential skill to enable Large Language Models (LLMs) to interact with the world. As tasks become more complex, they demand increasingly sophisticated and diverse reasoning capabiliti...
arxiv.org
December 3, 2025 at 12:42 AM
A thing they don't tell you when you travel for 16 of 48 hours to attend a sister's wedding over the weekend is that you don't get a weekend afterwards. Pretty tricky!
December 1, 2025 at 6:07 PM
Come join our department! It’s the best place to work in the world (he said with 0 bias).
We are hiring!!

University of Alberta in Canada has three faculty positions open this cycle: one in network security, one in theory, and one in robotics/computer vision/graphics.

Deadline: January 12, 2026

Happy to answer any questions about UofA or Canada academia!!

Please share!! 😁
December 1, 2025 at 2:45 AM