Posts for gbagcn

Post subject: Overview of deep learning
Experienced Forum User
Joined: 5/30/2005
Posts: 98
http://arxiv.org/abs/1404.7828 This is the only paper I have seen where citations make up over half its length (not even Wikipedia is this bad). Since these techniques have already been used to play atari games well maybe they will be used to make tas's one day. They would need to be an expert in both fields though and I am not sure if there is anyone like that.
Experienced Forum User
Joined: 5/30/2005
Posts: 98
Anty-Lemon wrote:
"That AI?" Do you have anything more specific than that?
http://www.engadget.com/2015/02/26/deepmind-atari-games-tests/ Sorry but I thought people here would have heard of this. Maybe this isn't as widely known as I thought. Also since the AI uses results from deep learning maybe that could be applied to tasing. I wonder if there is anyone who is both good with applying deep learning techniques and making tas's.
Experienced Forum User
Joined: 5/30/2005
Posts: 98
Ok fine I guess it is not dying. I guess I don't see as much discussion about tasing outside of here because everyone knows what it is by now. Also it seems like more work is being put into improving existing runs than creating new runs.
Experienced Forum User
Joined: 5/30/2005
Posts: 98
scrimpeh wrote:
Speedrunning in general is bigger than ever before right now. While I would honestly be hoping it would've made for a greater boost to the popularity of tasvideos, there are plenty of new users, submissions and progress on games being made. I'd certainly call the community far from "dying" at this point.
I seem to remember the forums being more active than they are now
Experienced Forum User
Joined: 5/30/2005
Posts: 98
hmm, I am pretty sure I made this topic but it got labeled as Guest and the first post got deleted since I haven't logged in for so long. That is an interesting forum glitch.
Experienced Forum User
Joined: 5/30/2005
Posts: 98
Patashu wrote:
I think there's some kind of cognitive bias for this. Summary of the bias is that you think there's a ton of content in the past because you can consume it all at once, but relatively little content in the present because you have to wait for it to trickle out, so you overestimate productivity of the past and underestimate productivity of the present/future. Also, people not coming as often doesn't indicate overall popularity is down. It's called churn, it exists in every field ever. The question is, are more new people becoming aware of TASes to make up for it, and I would likely say YES, if you look at the exponential growth of viewer base every GDQ for example, TASing interest is very closely related to that, especially ever since the GDQs started showing TASBot block. Another bias is that most people had more free time in the past than right now - as you get a job, study harder, etc. you have less time to watch and make TASes.
Actually when I am coming less often there should be more content since more time has passed between each visit
Post subject: Is the TAS community dying (No, it's not)
Experienced Forum User
Joined: 5/30/2005
Posts: 98
I used to come here more years ago but once most of the popular games were TASed I stopped coming as often. It looks like other people have been doing the same thing.
Post subject: Can that AI that plays atari games play better than a TAS
Experienced Forum User
Joined: 5/30/2005
Posts: 98
There was this technique the AI used when playing arkinoid where it kept the ball bouncing on the top of the screen that I had recognized from the arkinoid TAS that I had seen years earlier. That made me wonder how competitive it would be with TAS players.
Post subject: Why does google list the same links multiple times?
Experienced Forum User
Joined: 5/30/2005
Posts: 98
When I am going through google search results at around page 10 I start seeing links that I have already seen on a previous page. The same thing happens on youtube. Since google brags about finding thousands of search results on the top of the page you would think they wouldn't do this.
Experienced Forum User
Joined: 5/30/2005
Posts: 98
I remember reading something that said it was possible to beat this game without using the same item more than once throughout the whole game (excluding puzzle mode). Maybe you could make a TAS based on that.
Experienced Forum User
Joined: 5/30/2005
Posts: 98
Seems like an interesting game. Maybe you could make it so the scores on the balls wouldn't increase after a few seconds unless you sent the ball airborne. If not that I am sure there is something you could do to discourage doing nothing but keeping the platform flat.
Experienced Forum User
Joined: 5/30/2005
Posts: 98
Stumbling on Happiness One Jump Ahead: Challenging the Human Supremacy in Checkers The Singularity is Near A New Kind of Science These are some of my favorite books. You can find many previews of them on google books. I hope you like them.
Post subject: Whats the furthest you can get in this infinite mario game?
Experienced Forum User
Joined: 5/30/2005
Posts: 98
http://users.soe.ucsc.edu/~bweber/dokuwiki/doku.php?id=infinite_adaptive_mario I like this game better than the java game it is based on since it gets harder quicker. Also I found a glitch which causes enemies to completely disappear so don't post the level you get to if you encounter this glitch (with this glitch I got to a level over 1000). My best score is world 123?
Post subject: Shouldn't game emulators be called simulators instead?
Experienced Forum User
Joined: 5/30/2005
Posts: 98
From what I know about emulators outside of gaming, a emulation is a perfect simulation of something. Since game emulators are not perfect simulators of the hardware they are supposedly emulating shouldn't they be called simulators instead?
Post subject: Anyone have a Geometry Wars Galaxies save...
Experienced Forum User
Joined: 5/30/2005
Posts: 98
...with the lambda galaxy unlocked. For the unfamiliar with the game this is normally unlocked by connecting this game with the wii version.
Experienced Forum User
Joined: 5/30/2005
Posts: 98
Is there any way to change the button mapping. I don't really like the default
Post subject: Can you survive an infinite amount of time in a 3 vs 1 match
Experienced Forum User
Joined: 5/30/2005
Posts: 98
I have noticed in most multiplayer 1 vs 1 matches that if one player plays defensively enough he can survive for an infinite amount of time. What if it was 3 vs 1 instead? Lets say you take a 4 player game with everyone having access to TAS tools. Three of the players try to kill one player. That one players goal is simply to survive as long as possible (possibly an infinite amount of time). Is it possible to survive an infinite amount of time and if not what is the highest amount of frames that one player can survive for? When I thought of this idea SSB64 and Mario Kart 64 battle mode came to mind. What do you think the highest possible frames of survival are for these games.
Post subject: UK university develops massive games emulator
Experienced Forum User
Joined: 5/30/2005
Posts: 98
http://www.techradar.com/news/gaming/uk-uni-developing-massive-games-emulator-528487 Once its made I wonder if someone will add support for TASing.
Post subject: Idea for a ssb AI that can't be damaged
Experienced Forum User
Joined: 5/30/2005
Posts: 98
In ssb the thing that determines if you will be hit by an attack in the next few frames is your position relative to your opponents. This is also known as spacing. To define it even more precisely take the position of 2 characters in a match and draw a line between them. Now find the length of this line and its slope (this could also be refered to as angle and distance). Other things like what the character (the state of the character) is doing are also important too (is the character doing and attack and what direction are they facing). Now lets say there was an AI that knew all of the angle and distance positions that would result in getting hit in a few frames. If the AI moved in such a way to avoid these positions it would never get hit. I am interested in finding what these positions are but to do this I would need coordinate information that describes where both characters are at all times and I am not sure how to find this. Can anyone help me with this?
Experienced Forum User
Joined: 5/30/2005
Posts: 98
What about a pokemon AI that can play with any team. I wonder how hard this would be to make compared to a bomberman AI.
Post subject: What effect will quantum computers have on tasing?
Experienced Forum User
Joined: 5/30/2005
Posts: 98
Brute forcing to find the fastest possible tas should be much faster since you can try multiple button sequences at the same time.
Post subject: Creating a TAS doesn't seem very user friendly
Experienced Forum User
Joined: 5/30/2005
Posts: 98
The idea of creating a TAS sounds kind of interesting since I like to speedrun a few games. The problem is that when creating a TAS you have to make so many more choices. These choices include: 1. When to create a savestate 2. When to overwrite a savestate 3. When to load a savestate 4. If you choose to load a savestate which savestate to load 5. How many frames per second to play at and when to increase and decrease this amount All of these choices have alot more depth to them than just choosing what button to press like when playing a speedrun. Whats worse is that since these choices aren't really gameplay related so they aren't that fun to make. It really feels like there should be some way to automate these things. Also in addition to this you have to worry about things like manipulating luck, desynchs, and hex editing when creating a TAS.
Post subject: AI designed to play Ms. Pac-Man
Experienced Forum User
Joined: 5/30/2005
Posts: 98
http://www.robotworldnews.com/100389.php This is the first time I have ever heard of an AI being designed for a video game that wasn't designed by the original programmers. I wonder if any of these techniques could be used to make better TAS tools.
Post subject: How much space does 1 frame of a TAS movie file take up?
Experienced Forum User
Joined: 5/30/2005
Posts: 98
TAS movie files are made up of a bunch of button presses so I am guessing that it depends on the number of buttons the controller has. Since each controller button has 2 states (pressed, not pressed) that means each button should be able to be stored in a single bit since a bit has 2 states as well (0,1). The SNES controller has 12 buttons (d-pad counts as 4) so that means each frame should take up 12 bits right?
Experienced Forum User
Joined: 5/30/2005
Posts: 98
Ok just ignore that 3rd paragraph in my last post since I figured out I was doing some things wrong in that calculation. If you want to find the total amount of space every 256x256, 24 bit picture takes up first you multiply those numbers like I did earlier and get 1572864 bits. Then you need to figure out how many possible states those bits have and to find that do 2^1572864. Then take 1572864 x 2^1572864 and the answer should represent the number of bits all possible 256x256, 24 bit images take up. I am pretty sure that I did this right since this works for smaller file sizes as well.