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
Reposted by Matthew Guzdial
NEW: The first two years of Massachusetts' millionaire tax has raised $3 billion more than expected.

And rather than driving the rich away, IPS researchers found that the number of millionaires has *increased.*

Tax the rich. Greg Ryan in @bloomberg.com:
Millionaire Tax That Inspired Mamdani Fuels $5.7 Billion Haul in Massachusetts
A millionaire levy in Massachusetts that New York City mayoral frontrunner Zohran Mamdani holds up as a model for taxing the rich has generated $3 billion more in revenue than expected without forcing...
www.bloomberg.com
November 12, 2025 at 4:01 PM
Related but distinct from #AIIDE25 is the release of SIMA 2 from DeepMind. Will Douglas Heaven has a nice, measured writeup and kindly included some critical quotes from me in it.
www.technologyreview.com/2025/11/13/1...
Google DeepMind is using Gemini to train agents inside Goat Simulator 3
SIMA 2, which can figure out how to solve problems inside virtual worlds, could lead to more general-purpose agents and better robots.
www.technologyreview.com
November 13, 2025 at 5:57 PM
Follow along with us through Emily's live posting for #AIIDE25 Day 2!
Once again live posting AIIDE 2025 Day 2!
And we're back live with AIIDE 2025 day 2! We're starting off with a very exciting keynote from Luke Dicken @luked.bsky.social and Nadine Perez titled 'An Algorithm is Just the Start'.

Looking forward to getting some insight into the challenges in applying academic research to industry!
November 13, 2025 at 4:31 PM
Reposted by Matthew Guzdial
As of this morning, Starbucks workers across the country are officially ON STRIKE. And we're prepared for this to become the biggest and longest ULP strike in Starbucks history.

Say #NoContractNoCoffee with us: DON'T BUY STARBUCKS for the duration of our open-ended ULP strike! $SBUX
November 13, 2025 at 11:33 AM
Reposted by Matthew Guzdial
I am extremely honored and humbled to have been awarded a Test-of-Time award for my 2005 paper "From Linear Story Generation to Branching Story Graphs" with R. Michael Young
November 12, 2025 at 4:08 PM
Reposted by Matthew Guzdial
Hello everyone! Welcome to day 1 of AIIDE 2025, starting off with some opening remarks from Seth Cooper and Matthew Guzdial @matthewguz.bsky.social

Just like EXAG/INT, I'll be live posting throughout the presentations throughout AIIDE for those who cannot make it!
November 12, 2025 at 3:42 PM
Emily is again live-posting as we reach the first real day of #AIIDE25, follow along!
Hello everyone! Welcome to day 1 of AIIDE 2025, starting off with some opening remarks from Seth Cooper and Matthew Guzdial @matthewguz.bsky.social

Just like EXAG/INT, I'll be live posting throughout the presentations throughout AIIDE for those who cannot make it!
November 12, 2025 at 4:23 PM
Reposted by Matthew Guzdial
I would agree, and also say that proc gen *is* a form of AI, it's just that genAI has now become interchangeable with the term and nuance has been lost despite the fact that creative computing and algorithmic art has existed for decades
A big reason proc gen feels so different to AI in spite of superficial similarities is exactly this!!

Proc Gen systems reflect the interests and creative preferences of the tool makers, whereas AI is catch-all sludge that reflects no specific point of view. It’s not remotely the same!
In case anyone ever asks, I *have* written a material generator and a music generator for Eldritch 2, but that's not AI. That's me taking all the things I personally know about those things, and writing my own tools to make the process a little faster and easier for myself. It's still 100% me, baby.
November 12, 2025 at 9:02 AM
What other conference gives you the Aurora Borealis? Only #AIIDE25! (Photo credit: David Thue)
November 12, 2025 at 4:22 AM
Watching a NES running a neural network to detect hand drawn images LIVE. EXAG undefeated. #AIIDE25
November 11, 2025 at 10:21 PM
Reposted by Matthew Guzdial
Just want you to know that when I asked Geoff why he gave the silent treatment to an open letter written by his own "future of games" program (& co-signed by 3000+ industry peers), he replied:

"Because that's my show and nobody gets to tell me how I should use my platform". 💀💀💀
November 11, 2025 at 6:34 PM
Reposted by Matthew Guzdial
I was so so happy to give an invited talk today at EXAG, a workshop I co-founded over ten years ago! It was quite a personal talk about how research communities grow and change (and die sometimes), what generative AI means for our community today, and the value of weird, pointless research.
Starting off Day 2 of EXAG with Mike Cook's @mtrc.bsky.social amazingly titled keynote 'Being Weird at the End of the World'. Really looking forward to this talk!
November 11, 2025 at 5:57 PM
EXAG/INT Workshop Keynote from Mike Cook @mtrc.bsky.social with incredible slide design. Shout outs @bombsfall.bsky.social
November 11, 2025 at 4:41 PM
Follow along with Emily’s live posting of day two of the EXAG/INT workshop! #AIIDE25
EXAG/INT day 2 coverage is starting now!
Starting off Day 2 of EXAG with Mike Cook's @mtrc.bsky.social amazingly titled keynote 'Being Weird at the End of the World'. Really looking forward to this talk!
November 11, 2025 at 4:10 PM
Reposted by Matthew Guzdial
On my way to the 21st AAAI conference on AI and Interactive Digital Entertainment aiide.org

Easily one of my favorite conferences. Full of fun, passionate people

Stay tuned for some news
Artificial Intelligence and Interactive Digital Entertainment – Sponsored by the Association for the Advancement of Artificial Intelligence
aiide.org
November 11, 2025 at 2:20 PM
Reposted by Matthew Guzdial
Looking forward to seeing friends and colleagues in Edmonton this week for AIIDE! I'll be trying my hand at live tweeting the paper sessions at both the main conference and EXAG/INT, which I'll try to keep in this big reply thread (starting off tomorrow). @aiide.bsky.social @exag.bsky.social
November 10, 2025 at 12:39 AM
My fantastic PhD student Emily is live bskying #AIIDE25! Follow along to see summaries of all the fantastic @exag.bsky.social sessions today!
Looking forward to seeing friends and colleagues in Edmonton this week for AIIDE! I'll be trying my hand at live tweeting the paper sessions at both the main conference and EXAG/INT, which I'll try to keep in this big reply thread (starting off tomorrow). @aiide.bsky.social @exag.bsky.social
November 10, 2025 at 5:00 PM
Excited to be welcoming everyone to #AIIDE25 Here at the University of Alberta!
November 10, 2025 at 3:51 PM
Reposted by Matthew Guzdial
Just thinking again how Democrats asked people to accept short-term suffering in order to give space to prevent long-term, worse suffering and then said, "Thanks for suffering in the short-term. You'll get to suffer long-term too so that our Thanksgiving plans don't get disrupted."
November 10, 2025 at 1:18 PM
Reposted by Matthew Guzdial
AIIDE '25 proceedings are up! And the conference starts in two days. 🧳✈️
Vol. 21 No. 1 (2025): Twenty-First AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment | Proceedings of the AAAI Conference on Artificial Intelligence and Intera...
ojs.aaai.org
November 8, 2025 at 7:21 PM
Even outside of these results now (which are really interesting!), there’s no guarantee that LLMs will even retain this level of human likeness given that finding new sources of pure human text at scale is becoming increasingly difficult.
LLMs are now widely used in social science as stand-ins for humans—assuming they can produce realistic, human-like text

But... can they? We don’t actually know.

In our new study, we develop a Computational Turing Test.

And our findings are striking:
LLMs may be far less human-like than we think.🧵
Computational Turing Test Reveals Systematic Differences Between Human and AI Language
Large language models (LLMs) are increasingly used in the social sciences to simulate human behavior, based on the assumption that they can generate realistic, human-like text. Yet this assumption rem...
arxiv.org
November 7, 2025 at 3:05 PM
Reposted by Matthew Guzdial
LLMs are now widely used in social science as stand-ins for humans—assuming they can produce realistic, human-like text

But... can they? We don’t actually know.

In our new study, we develop a Computational Turing Test.

And our findings are striking:
LLMs may be far less human-like than we think.🧵
Computational Turing Test Reveals Systematic Differences Between Human and AI Language
Large language models (LLMs) are increasingly used in the social sciences to simulate human behavior, based on the assumption that they can generate realistic, human-like text. Yet this assumption rem...
arxiv.org
November 7, 2025 at 11:13 AM
Reposted by Matthew Guzdial
We’re stoked to bring this project back and cant wait to share it with y’all 🍃 #ttrpg #ttrpgs #thequietyear #CapCut
November 6, 2025 at 1:47 AM
Reposted by Matthew Guzdial
As I've always been saying, LLMs have no relation to Truth at all. They can answer questions, generate ideas on a theme, and many other amazing things, but they don't know the difference between true and false, opinion and hyperbole. These results support that.

www.nature.com/articles/s42...
Language models cannot reliably distinguish belief from knowledge and fact - Nature Machine Intelligence
Suzgun et al. find that current large language models cannot reliably distinguish between belief, knowledge and fact, raising concerns for their use in healthcare, law and journalism, where such disti...
www.nature.com
November 4, 2025 at 3:05 PM