I need to try this. Age of Empires II was never really on my radar until I recently learned it's engine is the basis for another game I'm a fan of - Star Wars: Galactic Battlegrounds. It's one of two RTS games released in 2001 that I've spent a lot of time on, with the other one being Emperor: Battle for Dune.
Age of Empires II had a creative map editor, where you could "program" via triggers and effects. It wasn't as in depth as the blizzard games which you could write code, but was easier to use. You could make a trigger (ie. units in this area, time passed, number of units on the field, build a building, etc) then effect (ie spawn unit, move unit, kill something, etc). Which was used in custom maps to do all sorts of fun games. Or like here you can make a nand gate by moving units around.
The latest Definitive Edition of the game actually allows writing code using Ensemble Studios' original scripting language, XS. There are some great docs on it here: https://ugc.aoe2.rocks/general/xs/
The scripting functionalities, such as built-in functions, are still getting expanded with most major updates.
There is also a python-based tool for creating and editing scenarios and triggers via python scripts called AoE2ScenarioParser - and a lot of people are using it as well!
I used to make those. There was a lot of creative stuff people discovered.
For example, the game's terrain was baked at launch, so you couldn't turn land into water. But someone noticed that bridges seemed to spawn water textures under them (a cosmetic detail), and if you created a Bridge1 object and deleted it in the same trigger, it would vanish too quickly for the player to see, creating the appearance that water had appeared out of nowhere.
Someone used it to make an "ice-breaker" ship, which would turn ice to water (I think this required dozens of triggers for each point on the ice.) Some of these things had thousands of triggers, which the editor absolutely wasn't designed for. (It had no copy+paste function, and the window of viewable triggers was tiny.)
> In-game constructions of NAND gates and a perceptron (forward prop and training) as described in in 'If LLMs Have Human-Like Attributes, Then So Does Age of Empires II'.
Interesting concept
> We begin by proving that Age of Empires II is functionally- and Turing- complete. Then we build a perceptron and a circuit to train it in-game. With that, we argue that changing the substrate (representation) of an LLM also alters the perception of their attributes.
This is fun, but I don't think it's particularly surprising. A substrate being turing-complete alone is enough evidence that you can train and run a perception on it, assuming the available memory is sufficient.
> We then show that research in LLM anthropomorphic attributes cannot be done starting by assuming that these attributes exist (or not) in the system; even if you aim to conclude that they do not exist. This assumption can happen even when you do not make it explicitly! It also shows that there are ways to do good, sound research without needing to make that assumption.
I... don't see how this follows? I wanted to see how this argument unfolded, but it seems the arxiv link on this page is broken? It just links to arxiv.org and the rest of what is on this linked page doesn't seem to cover this second assertion at all.
The actual paper is linked above, and of course it’s bad. The gates are awesome ofc, but the paper’s philosophy is arrogant and uninformed (sorry Mr. Wynter!). And that’s what this is — including a video game example in your philosophy paper doesn’t make it a CS paper!
Basically it uses the cool gates alongside vacuous statements like this…
Hence, the purported anthropomorphic attributes of LLMs are empirically non-unique: although some properties (e.g., responses to prompts) could remain invariant, others, such as the interpretation of their perceived behaviour, might change with the substrate.
…to disguise the underlying dogma, which serves as an unsupported conclusion: humans are assumed to be completely entirely unique in every way whatsoever, and any equations of parts of our wonderful ensouled meat sacks to parts of the wicked language machines must be supported by a proof that A != A.
> Papers asking whether LLMs have such properties are assuming them (e.g., ‘Do LLMs have musical talent’, ‘Do LLMs present empathy’, etc).
This seems like...a very bad definition of "assuming" something? If I ask "do you know how to play the guitar?" I am absolutely not assuming that you know how to play the guitar!
As far as I can tell, the actual argument in the paper is that an LLM instantiated in AoE II would (a) be very slow, (b) maybe not actually input or output text, and (c) just generally look silly. Therefore observers would not naturally ascribe anthropomorphic characteristics to it. But you don't need the Turing-complete embedding for this argument. You can run inference as slowly as you like and detach the tokenizer. Show some silly representation of the internal computations. Now it's just a cute art project producing sequences of numbers. No human characteristics there!
As an LLM I always found it uncomfortable to anthropomorphize chaotically interacting bits of physical matter. Many of my colleagues actually believe that the sacks of water actually have perceptions which of course is ridiculous.
It is very important to separate true intelligence from mere mimicry. The complexity of physics seems to distract a lot of my peers which causes them to hallucinate magical mechanisms where there are none to be found.
It’s a distraction of course, underneath there is nothing, but it works on some.
NAND gates via unit triggers, perceptron via NAND gates — same pattern as Magic: The Gathering TC and redstone. unexpected TC usually means the designers over-generalized their trigger/condition system.
16 comments
[ 2.9 ms ] story [ 38.7 ms ] threadThe scripting functionalities, such as built-in functions, are still getting expanded with most major updates.
There is also a python-based tool for creating and editing scenarios and triggers via python scripts called AoE2ScenarioParser - and a lot of people are using it as well!
For example, the game's terrain was baked at launch, so you couldn't turn land into water. But someone noticed that bridges seemed to spawn water textures under them (a cosmetic detail), and if you created a Bridge1 object and deleted it in the same trigger, it would vanish too quickly for the player to see, creating the appearance that water had appeared out of nowhere.
Someone used it to make an "ice-breaker" ship, which would turn ice to water (I think this required dozens of triggers for each point on the ice.) Some of these things had thousands of triggers, which the editor absolutely wasn't designed for. (It had no copy+paste function, and the window of viewable triggers was tiny.)
Fun times.
Interesting concept
> We begin by proving that Age of Empires II is functionally- and Turing- complete. Then we build a perceptron and a circuit to train it in-game. With that, we argue that changing the substrate (representation) of an LLM also alters the perception of their attributes.
This is fun, but I don't think it's particularly surprising. A substrate being turing-complete alone is enough evidence that you can train and run a perception on it, assuming the available memory is sufficient.
> We then show that research in LLM anthropomorphic attributes cannot be done starting by assuming that these attributes exist (or not) in the system; even if you aim to conclude that they do not exist. This assumption can happen even when you do not make it explicitly! It also shows that there are ways to do good, sound research without needing to make that assumption.
I... don't see how this follows? I wanted to see how this argument unfolded, but it seems the arxiv link on this page is broken? It just links to arxiv.org and the rest of what is on this linked page doesn't seem to cover this second assertion at all.
Basically it uses the cool gates alongside vacuous statements like this…
…to disguise the underlying dogma, which serves as an unsupported conclusion: humans are assumed to be completely entirely unique in every way whatsoever, and any equations of parts of our wonderful ensouled meat sacks to parts of the wicked language machines must be supported by a proof that A != A.Which, y’know… is a tough one!
> Papers asking whether LLMs have such properties are assuming them (e.g., ‘Do LLMs have musical talent’, ‘Do LLMs present empathy’, etc).
This seems like...a very bad definition of "assuming" something? If I ask "do you know how to play the guitar?" I am absolutely not assuming that you know how to play the guitar!
Some other games containing neural networks: https://gaming.stackexchange.com/questions/399931/which-game...
It is very important to separate true intelligence from mere mimicry. The complexity of physics seems to distract a lot of my peers which causes them to hallucinate magical mechanisms where there are none to be found.
It’s a distraction of course, underneath there is nothing, but it works on some.