There are several points that I disagree with here, but I'll try to be concise:
i will take them one by one.
Absolutes: "AI will never win"
There is evidence that shows AI beating humans at games. To say they will never with this game is not supported. It has not played this game.
well, no evidence available to your judgment, somehow. though i try to make a heartfelt case here, there is a good deal of practical evidence to support the assertion, some of which is alluded to in the article, including practicality and expense, plus the interesting issue raised in another reply, that by the time we get to the level of complexity in embodied intelligence supporting the multimodal temporal-aspirational aspects of 40k as a thing that people do, and try to do well, we get away from artificial and next to natural. i mean, we end up setting up systems according to general principles of natural systems, and then we spend the rest of the time trying to teach them the right ways and eventually get some use out of them. but really how is this any different than what we do with human beings already? by the time that some 'artificial' entity emerges completely free to determine its course in life, which is typically the very thing denied to robots and AIs, then that entity is a part of nature and the natural world and also thus deserving of moral recognition under law, and moral regard subject to moral sanction and so on...
Chess is a massive game that has been around for a massive amount of time. It is generally respected as a game that takes intelligence to win. There is incentive for someone to write an AI that beats the best in the world. 40k is by comparison, a wargame in infancy with less respect than chess. The incentive to throw resources at this game in the same manner as chess is greatly diminished. AI may not win, but it's because there isn't much of a demand to write one that will. (compared to chess)
there are AIs that play some parts of 40k, virtually, already. chess is easier.
and, you are right that nobody is gonna build a dedicated 40k gamer-bot,
your reasons are good ones, too.
Iterative and integer Thought
To think that AI is limited to "calculating move in advance" and "spaces" is a false and limiting assumption. Moves are combinations of threat bubbles, risk odds, and vectors instead of spaces. It's less infinite than it seems. Node based AI maps can easily solve these problems and make decisions without iterating through every possibility. As it plays, it learns bad moves. As an example of this on a smaller scale: Small Scale Mario Brothers AI
googles thing is a (for now considered to be) 'deep' network that runs through countless simulations in correction of error toward a global optimum condition, "You lose, I win." it is best understood as a dynamic complex system, or a specially arranged stack of them anyways.
The Best player
Chess has several methods for determining who the best player is. 40k is not as established or respected. AI Beat the best at chess. AI can therefore beat any human at that game. AI can beat humans, they are not all "Gary Kasparov"s of 40k.
yeah, i understand.
but i mean it this way.
an "AI" won't be beating me.
or any "me" type/level whatever that is i guess.
there will always be something left out, for example personal mortal investment as felt by a critter on the way to death, this sort of temporality that makes how we spend our sunday afternoons more or less important. if an artificial agent has this sense of its own mortality it will either have better things to do, like save the world for the robots!, or find its favorite pastimes in activities that better match its natural capacities, i dunno like crashing stock markets!
Despite humans generally thinking that AI can't imagine or innovate new thoughts, that's only partially true at this stage. Innovation, especially within a framework of rules, can and must be simulated in a learning AI. The random thought that says "try to move here and see if you are rewarded" is the type of thought that starts the learning machine learning. Simulation of innovation is as simple as attempting an action that there is no data on and randomly being successful at it.
for one thing, i believe that random is a stand in for 'we really don't know how it works'.
so, there is limited information in this type of language,
i mean it really only tells us where our knowledge ends.
but in general, rough terms, you are correct in your account.
error correction and feedback loops and so on... predictive coding models, freedom is random compositions of high level primitives, something like that...
Proxy by RTS
While the rules are made for an action based game instead of a turn based game. AI and humans competing within the same parameters will sometimes result in an AI victory. They do also win at turn based games. If AI can sometimes win at a game of similar type, then the evidence supporting a "humans will always win" hypothesis is greatly diminished. The evidence upon which you extrapolate is heavily weighted in favor of the opposition.
right. good. i see.
i maintain that 40k is more than a competitive framework with colored pixels.
given that 40k arises from the human (neck-)beard(y)ed mode of embodiment,
a different mode - an AI in a metal suit - won't be so good at it...
We do have that. Monkey and the Banana AI, Swarm AI, Conversational AI, etc. Specifically focused brain functions simulated with an inherent randomness built in to enable spontaneous new ideas. Inasmuch as a mouse is not likely to write an opera, neither is an AI of limited scope going to either. If AI can simulate a functions performed by a mouse or a bee, then it follows that it can be written to simulate any subset of functions of a species. 40k is a subset of human thought that does not rely on the entire function of the brain. Lacking a fully realized brain does not preclude AI from winning. While the AI may be written with the ability to select and play an MP3, it is just as useless to the 40k subset of required skills as the desire a living organism has to reproduce or eat. My point here, is that you don't need a complete brain to play 40k, so lacking a complete critter brain does not imply failure.
i disagree completely with everything in bold.
the desire to improve at 40k is akin to the becoming that marks the difference between mere existence and being.
not only do you need a complete brain,
cuz this is what 40k is in the beginning designed from, for, around and towards the improvement of,
you need a complete critter with mortal urgencies and all that neurophenomenological stuff that more analytic thinkers like to not analyze cuz it hurts deir wittle hedz.
A robot can be built to manipulate the pieces. I doubt there is any question in that. The AI that runs it doesn't need to be in the same device. The robot is a tool for the mind. The robot aspect doesn't really lend anything to the discussion of AI winning.
I probably have more, but these thought in no particular order I felt should be addressed a little bit. Time to work on my basement a little bit!
yeah, i mention this in the original post. a handphone Ai that plugs into a robotic assembly at the local game shop so you and throw down against your ANdroid pet!