Miles Cranmer@MilesCranmer
This is an insane paper and I love it
https://arxiv.org/abs/2605.31514
1:03 AM · Jun 6, 2026 · 538.7K Views
Yann LeCun Praises Paper Showing Neural Networks Run In Age Of Empires II
29510.7K1.2K6.1K543.7K
Sentiment
Many users praised the paper on neural networks in Age of Empires II that Yann LeCun highlighted for its clever arguments, while some dismissed it as sophistry or narrowminded fallacies.
Pos
65.8%
Neg
34.2%
40 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS16.8KBOOKMARKS165LIKES563RETWEETS24REPLIES10
ergelgru@ergelgru
@MilesCranmer Reminds me of one of the greats;
1dViews 16.8KLikes 563Bookmarks 165
Michael Levin@drmichaellevin
I think the basic issue in all such discussions is that these kinds of competencies aren't some kind of essential "human attributes". Measuring cognitive properties by humans is myopic and causes all kinds of pseudoproblems. The emerging field of diverse intelligence has better frameworks. A spectrum of highly variable intelligences, and lots of research on which kinds of architectures enable which kinds of patterns (of behavior, of computation, of physiology, etc.), is more useful for discussions of natural, artificial, and hybrid agents of varying provenance and composition. https://www.frontiersin.org/articles/10.3389/fnsys.2022.768201/full http://journal.frontiersin.org/article/10.3389/fpsyg.2016.00902/full?&field=&journalName=Frontiers_in_Psychology&id=196518 and more at https://drmichaellevin.org/publications/behavior.html, plus lots of other good labs.
1dViews 6.2KLikes 165Bookmarks 65
Brian 🔰@brianwut
@MilesCranmer burying the real result:
> Age of Empires II is ... Turing-complete.
1dViews 9.8KLikes 358Bookmarks 8
Zoe Wang@zoewangai
@MilesCranmer The study invalidates LLM research built on preset human-like characteristics. After building an in-game neural network in Age of Empires II, the authors find perceived AI anthropomorphism stems mainly from interface design instead of inherent properties. https://www.openread.academy/en/paper/reading?corpusId=549270147
1dViews 4.8KLikes 87Bookmarks 21
Oliver Traldi@olivertraldi
@MilesCranmer Very old idea https://en.wikipedia.org/wiki/China_brain
1dViews 2.3KLikes 31Bookmarks 11
twtfayta@twtfayta
@MilesCranmer as an entity in the greater boston area, I have also been suspicious of my human-like attributes
1dViews 3.9KLikes 80Bookmarks 1
D. Goldman@alc_anthro
@MilesCranmer Any Turing complete system can obviously run (with sufficient scale) an LLM. That's really all they've shown there. It's different from what an LLM is esp. post training. It is that training on human inputs and outputs that aligns it with the human mind.
1dViews 3.6KLikes 74Bookmarks 2
QuantNerd@FastTradeKing
It's absurdist, perhaps satirical. He's arguing against illogical reasoning of AI functionists, who claim that consciousness is an emergent property of LLMs, by showing that a large collection of digital coats could also exhibit the same emerging consciousness (this is proven, not theoretical).
The inference here is that one must use measurements to determine if there is truly emergent consciousness. For instance, computing the phi of an LLM (integrated information theory). I find this interesting, haven't seen it discussed before.
One could break the LLM in half (so to speak) at some particular link and determine how much does that constrain the past possible state of its inputs. Do this recursively for all of its weights (a big job but tractable) and measure its phi. High phi indicates emerging consciousness. Simple.
1dViews 3.2KLikes 30Bookmarks 6
Yann LeCun@ylecun
@MilesCranmer @skdh 😅😅😅
Miles Cranmer@MilesCranmer
This is an insane paper and I love it
https://arxiv.org/abs/2605.31514
8hViews 4.2KLikes 30Bookmarks 5
Mags@NeuroTechnoWtch
@MilesCranmer It’s not LEGO or Age of Empires…it’s a neural net. LLMs are made of neural nets. All neural nets are brain-like by design. This is idiotic. They should have done it with LEGO for it to make sense.
1dViews 4.1KLikes 18Bookmarks 3
gavin leech (Non-Reasoning)@gleech
@MilesCranmer https://faculty.ucr.edu/~eschwitz/SchwitzAbs/USAconscious.htm
1dViews 987Likes 18Bookmarks 4
JFK Files@read_jfk_files
very clever. this is basically a repackaging of Kant's theory that "the unity of aperception" is a priori in order for any perception to occur. this is the source of Anthropomorphic projection. in a sense, according to Kant the very first object of Anthropomorphization is always *yourself*. You have to imagine yourself as "a singular conscious thing that exists and can receive experiences" in order for you to have this thought in the first place. Got that? Anthropomorphism about yourself is *required* in order to have anthropormophism about external "beings."
this is why i've always said, if you think you're "into AI", like really deep into theory, but if you have not read Kant, then you don't know a damn thing about anything. You have no map, you have no blueprint, you don't know who you are or where you're going with AI. Without the Kantian framework you are blind and stumbling around in the dark.
1dViews 892Likes 14Bookmarks 4
Lon()@Lon
@NeuroTechnoWtch @MilesCranmer Legos arranged in a giant network structure to form the Death Star and basic blocks arranged in a giant network structure to form Age of Empires have just as much connectivity as a neural net and just as much right to be taken seriously and have their welfare considered.
1dViews 401Likes 14Bookmarks 2
jacket@keterosis
@ergelgru @MilesCranmer he cites this as the inspiration for the title
1dViews 1.2KLikes 24
Norrification 🇨🇦@Norrification
@MilesCranmer This is glorious. One of my favorite reads since an FMRI of a dead salmon. https://law.stanford.edu/2009/09/18/what-a-dead-salmon-reminds-us-about-fmri-analysis/
1dViews 286Likes 2Bookmarks 3
Ben Schulz@schulzb589
@MilesCranmer A modified "cha-cha" dance with an infinite memory tape is Turing Complete.
Computational universality allows for weird stuff.
One could run Doom on a brain organoid.
1dViews 756Likes 8Bookmarks 2
shadow - 2011 Remaster@foidvoid
@ergelgru @MilesCranmer https://www.youtube.com/watch?v=SxE5CwnmLCI lil b speaks of this
1dViews 713Likes 3Bookmarks 2
Wattus Maximus@WattusMaximus
@MilesCranmer AoE II is a bridge between Trebuchets and AI!
1dViews 1.1KLikes 11
Philip Winston@pbwinston
@MilesCranmer The answer to all of these “system A can’t possibly be conscious because just look at it” is it might have some “human-like qualities” but the quantity is 10^-100 so as far as the ability to do the NYT crossword puzzle it’s effectively zero.
1dViews 1.8KLikes 8
Joman 👺@jomangblandino
@MilesCranmer Hey, that’s exactly what I was thinking a while back. Why don’t we have more games that actually force LLM to stick to a strict regimen and consequences? The Sims series and SimCity are great examples, and as the AoE paper shows, there are others that do this really well.
1dViews 3.2KLikes 7
Load more posts