Jump to content

smiley

WFG Programming Team
  • Posts

    353
  • Joined

  • Days Won

    2

Posts posted by smiley

  1. Nobody knows why or how the engine got those weights. That's the point. Artifacts unseen by humans manifests in billions of nodes. The engineers at Amazon aren't stupid enough to not see the obvious source of bias. Its the non obvious ones which remains hidden in data sets. In fact, I would venture a decent guess that menstruation, pregnancies, general agreeableness, and overall aggressiveness influencing said output more so than most corporations being led by men. In which case, the data isn't necessarily wrong. Its just biology and body chemistry. The average hours worked also agrees with the sentiment.

    AI ethics exists not to vet data. Its to ensure the mathematical approach is actually socially fair as well. A military recruitment model would suggest soldiers to ideally be > 6ft straight males. Easy to see why. Easy to see its the most optimal. Not so easy to see if its the ethically right choice. Standard deviation would suggest that people not of those qualities might perform better as well.

    In statistics, stereotyping isn't exactly wrong. It's just probability.

    Then there is the more general question of whether or not reducing people to numbers is ethical.

  2. 19 hours ago, Sevda said:

    AI will not destroy the world unless the programmer decides to.

    Usually large neural nets are blackboxes whose mechanisms are not entirely understood by the programmers who make it. Big data tunes the weights. Said data cannot be completely understood by humans due to its scale. See the axed neural nets from various large projects. Most famously, the recruitment engine at Amazon turned out misogynistic and was subsequently pulled.

    Then again, skynet isn't a thing. But there is potential for negative impacts here which are unintended by the authors. We don't make AI with if statements anymore. We feed it with enormous quantities of data in hopes that it finds a pattern in there. The pattern might not always be ideal because bias exists which manifests as artifacts in data sets.

  3. 11 minutes ago, ChronA said:

    That's also why I suggest a 2 year period.

    Two years is too long without minor releases. Minor releases aren't really a thing since even small bug fixes can make versions incompatible. Minor releases should be compatible, else its a major release. There isn't a way to ship those changes without an actual release.

    • Like 2
    • Confused 1
  4. There is a pseudo-standard for getting heightmaps of size 512 (largest map size) from NASA Visible Earth topology maps if you are on Linux. All the random maps do it this way.

    /**
     * Heightmap image source:
     * Imagery by Jesse Allen, NASA's Earth Observatory,
     * using data from the General Bathymetric Chart of the Oceans (GEBCO)
     * produced by the British Oceanographic Data Centre.
     * https://visibleearth.nasa.gov/view.php?id=73934
     *
     * Licensing: Public Domain, https://visibleearth.nasa.gov/useterms.php
     *
     * The heightmap image is reproduced using:
     * wget https://eoimages.gsfc.nasa.gov/images/imagerecords/73000/73934/gebco_08_rev_elev_C1_grey_geo.tif
     * lat=25.574723; lon=50.65; width=1.5;
     * lat1=$(bc <<< ";scale=5;$lat-$width/2"); lon1=$(bc <<< ";scale=5;$lon+$width/2"); lat2=$(bc <<< ";scale=5;$lat+$width/2"); lon2=$(bc <<< ";scale=5;$lon-$width/2")
     * gdal_translate -projwin $lon2 $lat2 $lon1 $lat1 gebco_08_rev_elev_C1_grey_geo.tif bahrain.tif
     * convert bahrain.tif -resize 512 -contrast-stretch 0 bahrain.png
     * No further changes should be applied to the image to keep it easily interchangeable.
     */

    The actual map is split into 12 or so separate files IIRC and the above commands extracts Bahrain. The specific image file depends on the region you want to extract. Bahrain is in C1.

  5. 3 hours ago, Stan&#x60; said:

    What do you think of wow's idea to remove trireme and quinquereme boats and make them the same size as biremes and only available through upgrades. A bit like advanced and elite variations.

    Since the sizing is the biggest problem, it would be an improvement.

    Our maps are not large enough for accurate scaling of ships.

    • Like 1
    • Thanks 1
  6. Step 1: Write down an updated design document that reflects the current game. There was an attempt 2 or so years ago that failed.

    Step 2: Make it an open repository that anyone can propose changes to.

    Step 3: Add weighted voting to determine what gets accepted and what gets rejected. Someone who has been here for 5 years should obviously have more leverage than a new player. Might seem unfair, but knowing the culture and the community built around this is necessary to propose a change that's going to make most of that community happy.

    Step 4: ???

    Step 5: Profit

    • Like 2
    • Haha 1
  7. 1 hour ago, maroder said:

    So people who have the skills to actually do some changes are unfortunately better off time and motivation wise by just creating a mod,

    Either that or work on engine bugs and underlying code without actually focusing on gameplay. In other words, what developers usually do. You can count with your fingers the number of actual noticeable gameplay changes that have been committed in recent memory.

    No one wants to be that guy.

    • Like 2
  8. 9 hours ago, Gurken Khan said:

    I did read comments stating that some people wouldn't even read the closed forums

    I can attest that I wanted to write some stuff in there at least two times. One was regarding something random maps that someone wanted to know the feasibility of, and some balancing stuff + historical discussions. I am not DMing someone just to post a reply. And I imagine a lot of people feel the same way.

    If I have the chance, this game would be beyond recognition, that's how different the vision is. And the spectrum of stuff people like are probably just as vast. It's like three gears interlocking, they don't spin. Hence why such a thing as making a rigid design document is going to cause so much controversy that the accuracy fix in A22 would pale in comparison. But at the end of the day, whose game is this and whose it for are questions that will eventually need answers.

    I suspect balancing per se is so dead because of the amount of work involved in proposing changes. You can't simply say, "X op because I got steamrolled by someone using X yesterday". Subjective reasoning isn't good enough either.

  9. Without additional context, who knows.

    If you get sued, there is a greater than >90% chance that it will be considered fair use, because its non-profit, its used as reference and because screenshots have a legal precedence of not being considered under the same copyright as the software at least in the case of video games.

    1. The purpose and character of the use.
    2. The nature of the work that is used.
    3. The amount and substantiality of the work used in relation to the copyright owner’s work as a whole.
    4. The effect of the use on the market for the work.

    Fair use is subjective based on these 4 criteria.

    Regardless, online service providers fall under safe-harbor laws. Article 13 in EU also makes reasonable providence subjective based on the size of the platform, whether or not its non-profit.

    • Like 3
  10. I am pretty sure I have seen some of these arguments being used for adding stables.

    1 hour ago, Yekaterina said:

    Wastes space and nerfs cav strategies

    This not only contradicts the thesis of the thread but implies that building footprints and the fact that being scouted are actually side effects. What about house footprints? I am sure at times, that's an inconvenience too. The point isn't to create an autochess.

    Scouting introduces a new dynamic which is desired because the objective isn't linear gameplay. And I disagree that your opponent finding out that you are massing cav is actually a failure in game design. An equally valid stance is believing that not being able to scout stables gives an unfair advantage to the attacker and overpowers early cav strats. Ask any AoE2 player and they would probably agree with the latter.

    tl:dr; are these actually problems with stables?

    • Like 3
    • Thanks 5
  11. 15 hours ago, andy5995 said:

    I'd like to improve them if  possible

    Doesn't have anything to do with map generation, so out of scope for the map script.

    The reasoning presented back then was that square maps look more unnatural, sharp corners provide an unfair advantage, there is unequal distribution of area and they are around 27% larger than the equivalent circular map. To be honest, I don't really care and I highly doubt the original proposer cares either.

    • Thanks 1
  12. Possibly something similar to what Unity previously had (deprecated in favor of video player). MovieTextures. C++ would be doing the same, but for the GUI pages, it results in a nicer framework.

    Also, existing animations itself could be moved into C++ with a new paradigm that replicates CSS3. Instead of translating onTick, you just change the position and define a duration for all position changes. The actual translating would be internalised and not done in JS tick functions.

    Probably a nicer GUI system for page design.

    The CSS3 animation system can be realised in 0AD just via mods without engine changes, so maybe something fun for one of the interested GUI mod developers here.

  13. 9 minutes ago, real_tabasco_sauce said:

    I could probably make a diff for this, or at leat part of it. IDK about any GUI stuff, but changing the calculations looks pretty easy. My question is would anyone be interested?

    Submit the patch and find out. If your arguments are convincing, there is no reason why it won't be committed.

    • Thanks 1
×
×
  • Create New...