Post about your opion of what might happen in the future of TAS.
How many more or less rerecords will be used?
How much longer or shorter will the movie's take to make?
How much will the movie's change.
Joined: 8/14/2009
Posts: 4090
Location: The Netherlands
EDIT: changed my mind, it's probably just not going to be much different from what it is now. Maybe even more rerecords due to even bigger optimizing, but not that much.
http://www.youtube.com/Noxxa
<dwangoAC> This is a TAS (...). Not suitable for all audiences. May cause undesirable side-effects. May contain emulator abuse. Emulator may be abusive. This product contains glitches known to the state of California to cause egg defects.
<Masterjun> I'm just a guy arranging bits in a sequence which could potentially amuse other people looking at these bits
<adelikat> In Oregon Trail, I sacrificed my own family to save time. In Star trek, I killed helpless comrades in escape pods to save time. Here, I kill my allies to save time. I think I need help.
I also predict a huge increase of playarounds and glitchathons in the future, because, where's the fun in lazily running a program over a game?
A huge appeal of TASes is to see old TASes broken and new, exciting improvements. Once all of this is gone, there will need to be something else to fill in the slot.
In the future, all TASes will be made using bots that brute-force for the best possible input combinations.
wimbledonswirl wrote:
How many more or less rerecords will be used?
Several billions of rerecords.
More like 10100000 rerecords for brute force. If it were just a few billion rerecords, we could conceivably do brute force TASes now. I am pessimistic about brute force TASes myself. I think TASes in the future will be done with much more script assistance than now, and perhaps be done completely by a computer, but I do not think they will be provably the optimal solution.
Of course it depends on the complexity of the input (you think about DS runs using the touchscreen) and the movie length. Actually I wonder, how many rerecords would a brute forced SMB run take?
This has been brought up somewhere close to like a billion times already, but, for discussion sake -
A level of SMB is like 30 seconds long. That means 1800 frames. There are legitimately 6 buttons to worry about (ignore start and select).
So, to brute force that, you'd need (2^6)^1800 = (2.3485 * 10^108)^30
So, a bigger number than I can calculate. Don't count on true brute forcing ever happening. Heuristic algorithms are the way to go.
Sage advice from a friend of Jim: So put your tinfoil hat back in the closet, open your eyes to the truth, and realize that the government is in fact causing austismal cancer with it's 9/11 fluoride vaccinations of your water supply.
Brute force will -never- be viable for anything but the shortest term tasks.
Much much more likely are programs that are goal-driven, with an awareness of how the program works and when they have achieved it. Then you give them a list of goals and they optimize it genetically.
This has already been done by Bisquit (MM bot that drops powerups, lunar pool bot that plays the whole freaking game, solomon key goal driven bot)
Joined: 11/30/2008
Posts: 650
Location: a little city in the middle of nowhere
Personally, I think we are actually plateuing a bit. Brute forcing is not a viable option for any full length TAS, and heuristic algorithms have always been feasable, but generally unnecessary since it's usually easier to make the run by hand than to write an overly complicated program to do so. In most cases, bots are used to do extremely mechanical tasks that require little thinking and are simply boring to do by hand, or are used to do tasks that would require so many man hours to do, they would be infeasable for any real person.
So in answer to your questions, I think rerecords will increase on average, and that movies will take longer to make, but that's only because I think that more people will start TAsing more complicated, modern consoles.
... and heuristic algorithms have always been feasable, but generally unnecessary since it's usually easier to make the run by hand ....
And in some cases, a heuristic algorithm will fail. A heuristic algorithm can't really be programmed to handle novel happenings. For example, no algorithm could have been written to find this:
http://www.youtube.com/watch?v=rKZ7AunBKtM#t=3m26s
Sage advice from a friend of Jim: So put your tinfoil hat back in the closet, open your eyes to the truth, and realize that the government is in fact causing austismal cancer with it's 9/11 fluoride vaccinations of your water supply.
Only if quantum computers become a reality and affordable. Which might never happen.
(I can't say that finding the optimal input for a game is an NP problem, but it certainly sounds like it, and naive brute-forcing is very definitely an exponential algorithm, and you are not going to perform it on a run of any significant length in a reasonable time no matter how much computing power you have.)
TASing in the future is done solely by adelikat, DarkKobold, Cardboard, Tompa and Randil. Rerecords climb occasionally as high as 431922, but are mostly around 13000. It takes usually takes 2 weeks for adelikat and Randil to finish a TAS. 3 weeks for Cardboard and DarkKobold and 4 months for Tompa, since he likes to whore frames and restart his TASes. Movies will change a bit or a lot, depending on the game.
Today there are people who want to invest 10,000 rerecords to make a decent movie but there are also those that want to invest 200,000 rerecords to make a well-optimized and good movie (of the same game). It depends on how popular the game is, how much competition there is and on the person running the game and other factors. I'm not sure how this will change in the future... Maybe the first Super Mario Sunshine TAS will use 10,000 rerecords, then it will be more and more optimized from that point on. Given the popularity of the game there will probably runs using >200,000 rereocrds similar to the current SM64 TASes. I don't think that the techniques on how to create TASes will change much. TASing doesn't necessarily require knowledge on programming scripts/bots (that also depends on the game, since there are games where scripts/bots aren't needed).
How much longer or shorter will the movie's take to make?
As I already said, today there are those who run a game with 10,000 rerecords and ones that try their best using 200,000 rerecords. The latter will probably need more time, whereas the former can run games in a weekend's time. Personally I'm the type that tries to really optimize his movies, so I'm not making great progress on my WIPs (Work in progress movies).
Future emulators include Dolphin (gamecube) and other new-era consoles. Games for these consoles are more complicated (3-dimensional, acceleration, etc.) so they will be harder to optimize compared to ... some NES game maybe. In general, I don't think there will be much of a difference.
How much will the movie's change.
Most of the movies currently published are pretty optimized so they won't change much, I'd say (only if a new glitch discovery is made).
it's been said already that true brute-force bots will never happen, but I'd like to throw the numbers around again.
amaurea wrote:
More like 10100000 rerecords for brute force.
so far, the universe has existed for about 8*1060Planck times.
The visible universe is roughly (1026)3 meters large, which is equivalent to 10183 cube Planck lengths.[1]
If we make a best-case-assumtion that the universe can host a fully working computer at each planck length (it can't) and those computers can test a combination every planck time (they can't), we'll end up with a maximum of 10243 combinations tested since the big bang.[2]
An NES controller has 4 buttons and a dpad. Thus there are 16*9=144 possible controller states for each frame (in a one-player game without L+R). Brute Forcing 113 frames of NES input (just under 2 seconds) would require calculating 144113 > 10243 combinations, pretty much exceeding the physical computation power of our universe.[3]
TASes that are assisted by smart bots attempting a limited set of possibilities (i.e. short sequences for luck manipulation, greedy strategies, heuristics, ..) are possible, but their possibilities are only slowly increased by computational powers. 1 more frame brute forcing = 144 times the calculations = 14 years of computer innovation (according to the most popular misinterpretation of Moore's Law).
An actual brute-force-bot that'll accept a ROM and a winning-condition and produces a TAS? Not in this universe.
[1] We don't know how large the non-visible parts of the universe are, but that doesn't matter. Only the visible portion has the chance to communicate the results of their computations to us.
[2] Discounting relativity and quantum effects for ease of calculation.
[3] With a quantum computer, we only need sqrt(n) steps instead of n, thus doubling the number of frames calculated in the same amount of time/space; raising the limit of our universe's computational power to no more than 226 frames.
/edit: fixed math because something as complex as a digital dpad confused me.
Joined: 4/17/2010
Posts: 11475
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
as TAS EDITORs would be emdedded to many consoles, rerecord count will become ZERO XD
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
... and heuristic algorithms have always been feasable, but generally unnecessary since it's usually easier to make the run by hand ....
And in some cases, a heuristic algorithm will fail. A heuristic algorithm can't really be programmed to handle novel happenings. For example, no algorithm could have been written to find this:
http://www.youtube.com/watch?v=rKZ7AunBKtM#t=3m26s
Why not? An algorithm should only be fed the end state (a list of memory addresses and the values you want them to be) and the Rom. It should then "reverse" all functions from the ROM and use the new functions to determine from what states it is possible to reach the end state. Then determine from what state you can reach those states and so on until you get to the starting state.
In the future, all TASes will be made using bots that brute-force for the best possible input combinations.
wimbledonswirl wrote:
How many more or less rerecords will be used?
Several billions of rerecords.
More like 10100000 rerecords for brute force. If it were just a few billion rerecords, we could conceivably do brute force TASes now. I am pessimistic about brute force TASes myself. I think TASes in the future will be done with much more script assistance than now, and perhaps be done completely by a computer, but I do not think they will be provably the optimal solution.
In the future, all TASes will be made using bots that brute-force for the best possible input combinations.
wimbledonswirl wrote:
How many more or less rerecords will be used?
Several billions of rerecords.
More like 10100000 rerecords for brute force. If it were just a few billion rerecords, we could conceivably do brute force TASes now. I am pessimistic about brute force TASes myself. I think TASes in the future will be done with much more script assistance than now, and perhaps be done completely by a computer, but I do not think they will be provably the optimal solution.
How far in the future are you talking about?
67 years, 4 months, 3 days, 10 hours, 5 minutes, and 38 seconds.
Sage advice from a friend of Jim: So put your tinfoil hat back in the closet, open your eyes to the truth, and realize that the government is in fact causing austismal cancer with it's 9/11 fluoride vaccinations of your water supply.
Joined: 11/18/2006
Posts: 2426
Location: Back where I belong
DarkKobold wrote:
wimbledonswirl wrote:
amaurea wrote:
Mothrayas wrote:
In the future, all TASes will be made using bots that brute-force for the best possible input combinations.
wimbledonswirl wrote:
How many more or less rerecords will be used?
Several billions of rerecords.
More like 10100000 rerecords for brute force. If it were just a few billion rerecords, we could conceivably do brute force TASes now. I am pessimistic about brute force TASes myself. I think TASes in the future will be done with much more script assistance than now, and perhaps be done completely by a computer, but I do not think they will be provably the optimal solution.
How far in the future are you talking about?
67 years, 4 months, 3 days, 10 hours, 5 minutes, and 38 seconds.
Joined: 11/23/2010
Posts: 14
Location: Peotone, IL, USA
[ May I just say that, as a lurker poking into the boards, I believe that TAS creators/enthusiasts have among the most intriguing minds/conversations that I have had the privilege of witnessing? ]
www.NintendoLegend.com -- One gamer's quest to play and review every American-released NES game!
