No replies
Joined: 04/19/2009

Because time seems to be of the essence, and no one forewarned that Seth would be joining us for chat, many did not see this. On request, I am posting the log. Gaps are where I deleted quits or joins.

[7:27pm] SethCooperIRC joined the chat room.
[7:29pm] SethCooperIRC: hello

[7:36pm] SethCooperIRC: would anyone like to chat about newchapter?

[7:38pm] silverbergirc: Well, I haven't yet tried it, but the general consensus is that the wiggle seems to be problematic, as you may alreadt know.

[7:39pm] SethCooperIRC: well, it will behave differently than before
[7:41pm] SethCooperIRC: I think that a change in how wiggle behaves is something that will have to come with this update to the scientific code

[7:43pm] silverbergirc: Have you read the feedback regarding the speed issues? If not, I will post a link
[7:43pm] silverbergirc: http://fold.it/portal/node/996631

[7:44pm] SethCooperIRC: yes
[7:46pm] SethCooperIRC: this is why we have modified iterations to give more control over how long scripts will run for

[7:49pm] silverbergirc: The concern for players is that older machines might struggle to attain optimal scores.

[7:50pm] SethCooperIRC: hasn't it always been the case that a newer machine would provide some advantage?
[7:50pm] jeff101: I sent you some e-mails within the Foldit site.
[7:50pm] silverbergirc: According to some players, no.
[7:51pm] jeff101: garbage in -> garbage out
[7:53pm] SethCooperIRC: thanks jeff101
[7:53pm] SethCooperIRC: the Rosetta developers have some thoughts on potentially increasing the speed of things
[7:54pm] SethCooperIRC: it seems there are two things that can contribute to wiggle taking longer to run
[7:54pm] SethCooperIRC: one is that it is actually "slower", that is, scoring and doing the computations necessary take more time
[7:55pm] silverbergirc: Ok, I/we are grateful that you have joined us here, now. Unfortunately, the players that have more experience with the newchapter are not here at the moment.
[7:55pm] SethCooperIRC: the other is that it takes longer to converge, that is, each set of computations takes the same amount of time, but, it has to do more of them
[7:56pm] SethCooperIRC: yeah
[7:57pm] spvincent: Seth: are the parameters to WiggleSelected the same?
[7:57pm] spvincent: Well, just the one parameter really

[7:59pm] SethCooperIRC: which parameter?

[7:59pm] SethCooperIRC: wiggle precision?
[8:00pm] spvincent: Well, it the parameter that gives the number of iterations of some internal routine.
[8:01pm] SethCooperIRC: yes
[8:02pm] spvincent: Tx, And will one iteration in general be faster tha before?
[8:03pm] spvincent: structure.WiglleSelected (1) can take a long time now

[8:07pm] MikeCassidytoo: hi
[8:07pm] silverbergirc: hello mike
[8:08pm] MikeCassidytoo: So Seth newchapter: when are you guys thinking of bringing it online?
[8:08pm] SethCooperIRC: right, now in newchapter an iteration should be relatively short
[8:08pm] SethCooperIRC: next Tuesdat
[8:08pm] silverbergirc left the chat room. (Quit: http://www.mibbit.com ajax IRC Client)
[8:08pm] SethCooperIRC: Tuesday

[8:09pm] silverbergirc joined the chat room.
[8:09pm] silverbergirc was promoted to operator by ChanServ.
[8:09pm] Susume2: howdy
[8:09pm] MikeCassidytoo: It is extremely slow; I feel when I play it I am basically being asked to be Boinc but choosing Scripts
[8:10pm] SethCooperIRC: can you elaborate on what you mean by slow?
[8:11pm] MikeCassidytoo: Hold on I need to help my mother in law stand
[8:11pm] auntdeen2: hi Seth
[8:11pm] Susume2: getting the same work done (e.g. wiggling to stability) takes 2-3 times as long
[8:11pm] auntdeen2: Tuesday?
[8:11pm] spvincent: Takes a long time! Particularly with multiple bands

[8:12pm] auntdeen2: is there a big rush for this?
[8:12pm] Susume2: the speed issue is separate from the issue of adjusting to the new meaning of an iteration
[8:12pm] SethCooperIRC: yes, we are in kind of a rush for this
[8:12pm] auntdeen2: why?
[8:12pm] SethCooperIRC: we have somewhat of a backlog of puzzles that are waiting on this to release
[8:13pm] spvincent: Unless one iteration in newchapters does significantly less work than currently, the proposed mechanism for implementing wiggle will take forever.
[8:13pm] auntdeen2: you can't release them in our current client? while players figure out how to handle all the scripts?
[8:14pm] MikeCassidytoo: We were hoping to test it against Devprev and maybe Main to understand how it will effect [aly
[8:14pm] SethCooperIRC: one iteration in newchapter currently does do less work than one in main
[8:15pm] MikeCassidytoo: You can not issue it as a Devprev and leave Main alone?
[8:15pm] Susume2: yes, it takes as many as 30-80 iterations of wiggle to do what 2 iteartions currently does
[8:15pm] auntdeen2: that would be ideal, so that all the bugs can be worked out - the speed issues - and the script conversions
[8:15pm] Susume2: the iterations go by quickly, but the sum goes by slowly
[8:16pm] spvincent: tx susume: that's what I wanted to know

[8:18pm] SethCooperIRC: since this has a large number of fixes to improve the scientific accuracy of the game, we have a number of puzzles that are waiting on this to get posted
[8:18pm] SethCooperIRC: so that's why we would like to get it released as soon as pozzible

[8:18pm] auntdeen2: Iwe understand that, Seth - but why can't they be released to our current client?
[8:18pm] auntdeen2: this is a radical change
[8:19pm] auntdeen2: it will take time for players to get used to - and there are hundreds of hours of script adaptation to do

[8:20pm] auntdeen2: I am sure that you all are very excited about this
[8:20pm] auntdeen2: and would like to get it going ASAP
[8:20pm] auntdeen2: but players need time to adjust
[8:21pm] SethCooperIRC: well, there are known problems with the scoring as discussed in the blog post
[8:21pm] auntdeen2: so the new client will score better?
[8:21pm] SethCooperIRC: so, rather than continuing to use a known inaccurate scoring function, we'd like to start using the fixed one
[8:21pm] auntdeen2: or more points, as it were
[8:21pm] Susume2: we need some time to test the looping wiggle in some of the more common scripts; sadly the blue fuse with the looping wiggle crashes the game
[8:22pm] SethCooperIRC: not necessarily more points, but more accurate points
[8:22pm] auntdeen2: sigh
[8:22pm] MikeCassidytoo: AD: its not about points but about scoring leading to accurate proteins

[8:22pm] auntdeen2: Seth - with all due respect
[8:22pm] NickyCGS: So far I haven't encountered a crash with that script and your uploaded solution Susume. I'm trying it with a few of my own now as well.
[8:22pm] SethCooperIRC: certainly if there are crashes please report them
[8:22pm] auntdeen2: what will this client add to the player's experience?
[8:22pm] vakobo: Linux client is very unstable
[8:22pm] auntdeen2: and the "fun"?
[8:23pm] Susume2: I hear you Nicky, thx

[8:23pm] SethCooperIRC: we have been trying to reproduce the reported crashes with limited success so far
[8:23pm] Susume2: given the small number of people testing newhapter, the crashes are affecting a sizeable fraction of them

[8:24pm] SethCooperIRC: of course the fun and the player experience are very important
[8:24pm] SethCooperIRC: but the scientific accuracy of the results is as well
[8:24pm] MikeCassidytoo: Will there be help in bring old scripts up to date; there are a lot!?
[8:25pm] auntdeen2: Mike - there are thousands
[8:25pm] SethCooperIRC: so, these are things that have to be balanced to some extent
[8:25pm] auntdeen2: Seth - we have not had the accuracy for 5.5 years
[8:25pm] drjr: vakobo is right about linux version
[8:25pm] auntdeen2: I can't imagine that it has become an emergency to do it before it is vetted
[8:26pm] auntdeen2: have the foldit backers said "get more accuracy or we withdraw the funding"?
[8:27pm] SethCooperIRC: this is part of the process - we need to make changes to keep the game up to date
[8:27pm] spvincent: Are these changes being rolled out in Rosetta also?
[8:28pm] auntdeen2: we know there should be changes - some we have asked for - for years
[8:28pm] MikeCassidytoo: I agree Seth BUT unless all the NewChapter puzzles are small for a while it will be very very daunting and boring to play a puzzle that takes 2 hours to do an beginning shek& wiggle
[8:28pm] auntdeen2: but this is too much rush for something that throws a new bug everyday
[8:28pm] auntdeen2: and is slow as molasses and needs big fast hot computer
[8:29pm] auntdeen2: and thousands of scripts adapted
[8:29pm] SethCooperIRC: yes, these changes are part of Rosetta and are already being used in Rosetta@home
[8:29pm] auntdeen2: next Tuesday is way too soon
[8:30pm] jeff101: have you measued performance declines in Rosetta@home? what declines have you noticed?
[8:31pm] spvincent: The last update to Rosetta was on October 10th.

[8:31pm] SethCooperIRC: well, the reason it went into R@h is that it's more accurate

[8:32pm] jeff101: but how much less troughput do you get with the better accuracy?
[8:33pm] MikeCassidytoo: Seth: Can you put the puzzles we asked for in both NewChapter and DevPrev so we can get a real feel for whats going on?
[8:33pm] drjr: Woud it be possible to have a toggle switch to perform initial optimizations with the old function then wiggle with the new?
[8:34pm] SethCooperIRC: yes, I'm planning on putting up the requested puzzles
[8:34pm] MikeCassidytoo: Thank you; that will help a lot
[8:34pm] auntdeen2: will help with the scripts, also
[8:34pm] Wbertro: NickyCGS: update this feedback for you
[8:34pm] Wbertro: http://fold.it/portal/node/996625
[8:35pm] NickyCGS: Thank you very much Bertro.
[8:36pm] SethCooperIRC: and of course once newchapter is released, we'll continue to work on it and refine it
[8:37pm] spvincent: Are you going to provide scripers with control over what goes into the Undo buffer?
[8:37pm] SethCooperIRC: we have discussed that
[8:37pm] SethCooperIRC: what is the use case for that?
[8:37pm] auntdeen2: the biggest issue is whether or not players with average or older computers will still be able to play
[8:37pm] SethCooperIRC: can you use quicksaves?
[8:38pm] spvincent: When running a script using the proposed new mechanism for scripting. The Undo graph will quickly get full of intermediate wiggles

[8:39pm] SethCooperIRC: is that by itself a problem or does it prevent some use of the undo graph?
[8:39pm] spvincent: Don't see how quicksaves apply in this situation

[8:41pm] spvincent: It's bad in the sense that when running a script and seeing a pose you like, you may be unable to recover it because the Undo Graph is being filled up with the results of uninteresting intermediate wiggles.
[8:41pm] SethCooperIRC: i see
[8:42pm] auntdeen2: I am trying to open my newchapter client, I didn't do the intermediate update, but just allowed this one - I am getting error that tells me to reinstall

[8:44pm] MikeCassidytoo: Yes Ad but just do it again; after first being zapped I was able to get on
[8:44pm] SethCooperIRC: we may be able to add scripting undo support
[8:44pm] auntdeen2: 6 times already, Mike
[8:45pm] spvincent: good: that would be welcome.
[8:46pm] smilingoneIRC: ok

[8:47pm] auntdeen2: spvincent - does the graph do the same thing when hand wiggling?
[8:48pm] smilingoneIRC: I've noticed that every update to the game makes me run slower in general. I dread updating to newchapter
[8:48pm] smilingoneIRC: after all I've heard about it.
[8:49pm] spvincent: ad: I'm referring to the case of a script where wiggle is modified in the way suggested by jflat. It's in the future.
[8:49pm] smilingoneIRC: But, if it makes the protein designs more accurate I can see the need.
[8:50pm] auntdeen2: thx, sp

[8:50pm] spvincent: I'm not looking forward to changing all the scripts.
[8:51pm] SethCooperIRC: maybe there's some kind of tool we could provide that would help convert the scripts
[8:51pm] spvincent: I think they're going to have to be modified in different ways. One size does not fit all.
[8:52pm] MikeCassidytoo: I had no other clients running AD
[8:53pm] auntdeen2: teams without currently active scriptwriters will have a hard time
[8:53pm] spvincent: Perhaps something that internally multiples the number of wiggles by 10 or so.
[8:53pm] auntdeen2: Mike - I duplicated and downloaded the update, was the only way
[8:54pm] MikeCassidytoo: How about a library script by you guys we could just add to all the scripts?
[8:54pm] MikeCassidytoo: We add a single line call to the library for it
[8:55pm] SethCooperIRC: maybe something that just does a search/replace on the code?
[8:56pm] SethCooperIRC: in the script editor
[8:56pm] MikeCassidytoo: A hack we can add now quickly and then over time refine as we understand/see whats going on
[8:56pm] spvincent: Not sure if that would work. Sometimes you wqnt to let a wiggle run to the bitter end: other times, such as at low CI, it doesn't matter so much

[9:00pm] MikeCassidytoo: Suddenly quiet??
[9:00pm] MikeCassidytoo: or did I disconnect?
[9:00pm] auntdeen2: I'm running blue fuse over & over, fascinating
[9:00pm] MikeCassidytoo: hehe
[9:01pm] auntdeen2: local wiggle doesn't seem to do anything (zero zip nada)
[9:01pm] MikeCassidytoo: On a new puzzle or one worked on before?
[9:01pm] auntdeen2: just did a rebuild on the ED I had played with
[9:03pm] auntdeen2: Seth - you still here?

[9:03pm] MikeCassidytoo: He is logged on
[9:04pm] Susume2: search and replace in scripts would be bad - somtimes you expect teh wiggle to lower the score and don't want to test for a minimum gain
[9:04pm] auntdeen2: I asked a serious question about "fun" before
[9:04pm] auntdeen2: we play the game to contribute to science - and the community - and for fun
[9:04pm] spvincent joined the chat room.
[9:05pm] SethCooperIRC: hello
[9:05pm] SethCooperIRC: i do have to go pretty soon
[9:05pm] auntdeen2: …and "fun"?

[9:06pm] MikeCassidytoo: Because we did not know you jumping in tonight some important people were not here; it would be good for you to hear from them

[9:07pm] MikeCassidytoo: Timo for one
[9:07pm] SethCooperIRC: yes, we are interested in the game being fun and engaging
[9:07pm] SethCooperIRC: this change may cause the style of gameplay to change somewhat
[9:08pm] MikeCassidytoo: That is true, Seth
[9:08pm] auntdeen2: then please don't rush this - if for no other reason than we need time to get used to it - identify the bugs - and be able to help newcomers
[9:08pm] SethCooperIRC: it may emphasize other aspects of gameplay than running wiggle to completion
[9:08pm] auntdeen2: like ?
[9:08pm] MikeCassidytoo: I wonder if this will effect the Turorials......
[9:09pm] Susume2: it is common practice to compare several rebuilds or tweaks by shaking and wiggling each one to stability adn keeping the best

[9:10pm] SethCooperIRC: new strategies may emerge
[9:10pm] jmbrownlee333: i just wanna state the obvious. this game is nothing if it can not adequately predict the native state. just a vote of support for the progress, even if it will cause hardship.
[9:10pm] auntdeen2: please explain, Seth - we are all just trying to absorb the implications
[9:11pm] MikeCassidytoo: Is it possible to put the tutorials online in NewChapter so I could play them?
[9:11pm] auntdeen2: jm - no one disagrees with better accuracy at all !
[9:11pm] jmbrownlee333: i know. as i said . stating the obvious

[9:12pm] auntdeen2: a change of this magnitude needs to be vetted as much as possible, the implications sorted out - and I will stand by saying that next Tuesday is simply too early
[9:13pm] SethCooperIRC: for example, a different way of evaluating the quality of rebuild or tweak
[9:13pm] auntdeen2: ah ha - 6 times running blue fuse stabilizes
[9:13pm] spvincent: I agree: a rather slower introduction would have been better
[9:14pm] jeff101: I think when newchapter starts, we will adapt. We will find which old scripts suddenly work best and which old favorites to revise.
[9:14pm] auntdeen2: jeff - we have done that many times
[9:14pm] jeff101: One question to ask is "is this a good time to deal with such a big change?"
[9:14pm] jeff101: in the middle of CASP would not be a good time
[9:14pm] jeff101: maybe now is a good time
[9:15pm] auntdeen2: and we can do it again - proveded that some do not stop playing because of their computers
[9:15pm] MikeCassidytoo: In the middle of CASP would be crazy
[9:15pm] auntdeen2: it's actually ideal timing
[9:15pm] SethCooperIRC: that's true, the more time we have before CASP the better
[9:15pm] jeff101: maybe you all could give us a break and give us the smallest puzzles first, or longer deadlines
[9:16pm] auntdeen2: jeff - targets are set by predictioncenter, not foldit
[9:16pm] auntdeen2: they set the deadlines
[9:16pm] auntdeen2: ah sorry - you mean in the new client
[9:17pm] jeff101: I mean the standard Science Puzzles we will do this spring
[9:18pm] SethCooperIRC: well, I've got to go now
[9:18pm] SethCooperIRC: thanks for the feedback
[9:18pm] NickyCGSirc: MikeCassidy, you can already access the tutorials on Newchapter.
[9:18pm] jeff101: and give us some time before the next PBS show about Foldit
[9:18pm] auntdeen2: thanks for stopping by, Seth
[9:18pm] jeff101: wouldn't want a flood of new players while a big change is going on
[9:18pm] SethCooperIRC left the chat room. (Quit: http://www.mibbit.com ajax IRC Client)
[9:19pm] Susume2: we may need a flood to make up for the attrition

Sitemap

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons