TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Algorithms that select which algorithm should play which game

63 pointsby togeliusalmost 9 years ago

10 comments

iandanforthalmost 9 years ago
While interesting, this research is not on the path to &quot;true&quot; artificial intelligence. It may solve playing a majority of video games given enough effort, but that is not the author&#x27;s stated goal. Why do I think this?<p>The representations of the various algorithms are not unified. The answer to the question &quot;what algorithm lies halfway between JinJerry and YOLOBOT&quot; is difficult to answer. This is because they are discrete solvers who&#x27;s implementations cannot be seamlessly blended. It is only at the level of their decision outputs that they share a common language.<p>The strength of natural intelligence derive partly from the fact that all strategies are implemented using the same components. Strategies can be described as sparse activation patterns of a neural substrate. Any strategy can be added to, subtracted from, or otherwise combined with another because the activations have a common representation language. The choice is never &quot;either&#x2F;or&quot; for natural intelligence but &quot;how much of which?&quot;<p>This problem is also found in the meta-selection problem. The decision tree used to choose a component algorithm (even if applied repeatedly throughout the course of a game) doesn&#x27;t allow for strategic blending, and itself is not implemented in a language common to the strategies.<p>In contrast, biological strategies are selected by competition between signals from many lower level systems and top down control and predictions. Ultimately a decision such as &quot;fight or flight&quot; comes down to how strongly a sparse set of neurons if firing and which motor paths are suppressed and which are activated. Because both the strategies and the evaluation of those strategies are implemented in a common substrate, you can blend, adapt, compare and update all aspects of the system. This is crucial to the adaptability and speed of natural intelligence.<p>Edit: I should note I am in full agreement with the author that simulated environments (of which video games are an example) are going to be essential for the creation of &quot;true&quot; AI. My comments relate specifically to the line of inquiry in the papers described.
评论 #12264688 未加载
评论 #12265244 未加载
daveguyalmost 9 years ago
This is a great post. (EDIT: Great Blog all around!). The grid of how different algorithms perform is particularly nice. I would love to see how that grid with new algorithms (specifically those new algorithms that make use of multiple algorithms).<p>Also, the name hyper-heuristics sounds like a recently made up term to stand out in search queries. The more long established name for it has been meta-heuristics (which is included as a keyword in the paper linked as &quot;hyper-heuristics&quot;). Meta because they are heuristics of algorithm selection rather than heuristics of solving a specific problem and heuristics because, as iandanforth mentioned, this is not a low dimension continuous differentiable problem space that lends itself to an optimal solution.
评论 #12267951 未加载
nickpsecurityalmost 9 years ago
Glad you submitted this as I&#x27;ve needed an update of game AI field. A bit different than it was in my day but still shows what I argued all along: hybrid methods will be the best. Far game generation, I agree that generating many new, unpredictable experiences for the algorithm is ideal. I also agree it&#x27;s within computational reach. Doesn&#x27;t necessarily take AI, though.<p>For instance, much of that could be done with declarative, dynamic programming with templates, aspects, or constraint solvers. Idea being you create objects with relationships, attributes, and constraints. Let&#x27;s look at an item object. It might have physical properties such as shape, size, movement speed, acceleration speed, ability to phase in&#x2F;out, invisibility, and area of effect. It might have reactive properties where any of that changes in a specific way upon physical interaction, game event, or global setting. It might have effects on players that changing any existing attribute&#x27;s value, remove one, or add one. It might cause a pre-registered event in game or for player with specific or random values input. Any of this can be programed using available languages as taking inputs, performing a computation on them, and producing output. Declarative aspect means that, after each is created with constraints or types, inference algorithms (even if-then&#x27;s if you&#x27;re masochist) can produce an imperative implementation that lines them all up properly for an actual game.<p>I&#x27;m not sure what the difficulty will be to do such things for an entire game rather than one item. However, one item on a map interacting with global &amp; one, player&#x27;s state trying to do all of the above might make a nice testbed. Then a number of items. Then a number of players. Whatever method works easily to solve it and generate efficient code gets used when approach is expanded to apply a similar range of attributes to player, map, NPC&#x27;s, or even passage of time itself.<p>Just some thoughts on that. Haven&#x27;t read your papers about specific algorithms yet. Did download Panorama for later reading. So, is anything close to what I described for generating a game engine from descriptions of parts in a way I described items? I think that, once enough stuff was loaded in, would generate a combinatorial explosion like wannabe AI&#x27;s have never seen (and couldn&#x27;t cheat easily).
评论 #12268121 未加载
oslavonikalmost 9 years ago
Great post.<p>OT: I accidentally swipe left&#x2F;right <i>every</i> time I&#x27;m on blogspot, taking me to the previous&#x2F;next post. Maybe I&#x27;m just fat-fingering all around, but this is a horrible UX.
评论 #12267956 未加载
mikekalmost 9 years ago
For those of you interested in this general area, I suggest taking a look at General Game Playing. <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;General_game_playing" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;General_game_playing</a><p>Disclaimer: My PhD advisor was the originator of this idea.
评论 #12274649 未加载
daveguyalmost 9 years ago
On the grid graph comparison of how different algorithms perform on different games, two questions:<p>1) What is the source data for that plot?<p>2) You specify &quot;lighter = better&quot;, but how are they normalized across games and algorithms? How is better and worse quantified to get a &quot;lightness&quot;?<p>Edit: Found #2 in the second paper. Still don&#x27;t know what 25 wins is white and 0 wins is black means? How do you &quot;win&quot; some of these?<p>Two papers are here:<p><a href="http:&#x2F;&#x2F;julian.togelius.com&#x2F;Mendes2016HyperHeuristic.pdf" rel="nofollow">http:&#x2F;&#x2F;julian.togelius.com&#x2F;Mendes2016HyperHeuristic.pdf</a><p><a href="http:&#x2F;&#x2F;julian.togelius.com&#x2F;Bontrager2016Matching.pdf" rel="nofollow">http:&#x2F;&#x2F;julian.togelius.com&#x2F;Bontrager2016Matching.pdf</a>
评论 #12267883 未加载
mccourtalmost 9 years ago
Great article and great paper. Similar set of questions that get asked as part of AutoML, including the idea of hyper-heuristics, but in the AI community instead of data science. Thanks for the insights.
causalmachinealmost 9 years ago
To summerize iandanforth&#x27;s longueur:<p>Togelius&#x27; work ignores a core paradigm, of which we are aware of the human brain; neuroplasticity.<p>If one of togelius&#x27; solvers fails, all bets are off, the artilect is lobotomized.<p>See deepmind&#x27;s &quot;self-encoding&quot; work: <a href="http:&#x2F;&#x2F;www-personal.umich.edu&#x2F;~reedscot&#x2F;iclr_project.html" rel="nofollow">http:&#x2F;&#x2F;www-personal.umich.edu&#x2F;~reedscot&#x2F;iclr_project.html</a>
fitzwatermellowalmost 9 years ago
Link to paper needs to be fixed.<p>Thanks togelius for the survey of the current state of the art around GVGAI! It will be interesting to see how adversarial methods influence future game design. Is a perfectly &quot;instinctual&quot; game possible? One that can never be beaten by any machine but that even a two-year old can master immediately?
评论 #12263221 未加载
fisherjeffalmost 9 years ago
Looking forward to the algorithm that selects which of these algorithms to use