March 28, 2005
I am What I Can Memorize.
This is my final project plan:
I am more than what I'm saying I am - it's less believable. I'm more true when my POV is lent to viewers, by which I've been seeing the beings where I live in - physical surroundings, people with whatever relationships to me etc, and the way my memory re-arrange the time/space. Much of my memory are fortunately well stored into my visual journals - oh, those footages.
So, my project will be a piece of hybrid media to be interacted with. I'll be edited in the footages I shot both at home and here. I'll be talking about myself. I'll be in the text I'm writing. I'll mingle them together and render the "viewers" a mouse and screen at last.
(BTW, the great enjoyment of writing visual diary is at the cost of spending a lot of time on solving intracable problems while dealing with hundreds of hours of footages. The Rule Of Play is: You lagged yourself yesterday on doing that - pilfering shooting, re-recording(if possible), editing daily, being carful of the stumbling blocks by your feet while shooting, moisture proofing, capturing, editting, rendering....etc., etc - you risk yourself of lossing the fun of it forever - esp. in the case that you've not a chance to take Peggy's IW course!)
March 23, 2005
Games, Cinema, Genre, etc.
Here's a brief summary of my comment in the writing class today:
Corporate/industrial similarity between movie industry and game industry:
--it seems that EA (and perhaps the other large game studios) occupy the risk-averse position that Hollywood studios occupy, leaving artistic innovation to exist in the independent film arena (whether or not financed by large studios); does an independent game development movement exist?
Conceptual connections: narrative vs. spectacle
--the study of the silent era of cinema reveals that narrative/fictional films did not dominate the landscape as it now does (and has since the 20s); Tom Gunning writes of what he calls the "cinema of attractions" where the pleasures derive from cinema as spectacle less than any development of a story; in this viewing context the image is an object observed more than a world inviting the viewer into
--spectacle still appears in cinema, particularly in the Hollywood summer blockbuster, wherein action sequences exist largely for their thrill ride quality rather than any narrative function
--games offer some sort of triangulation of three pleasures: narrative, spectacle, and interactive play (this might be the capacity to explore a world, some sort of competitive/achievement pleasure, some sort of pleasure in the physicality of the gameplay experience); different games (obviously) approach these pleasures with varying degrees of engagement, either by virtue of their type of game or their combination of effects (exploration + cut scenes + graphics, etc.)
--for further reading on silent cinema and spectacle, see Tom Gunning; for further thoughts on spectacle replacing narrative in contemporary cinema and new media, see Andrew Darley "Visual Digital Culture" (I think that's the title; I might be able to dig it out and show you Wed morning)
And finally, on genre:
--genre in terms of video games often gets discussed as FPS, RPG, RTS, side-scroller, puzzle game, etc., whereas genre in film generally indicates a type of narrative structure along with certain conventions, themes and iconography. (Genre in literature is often form specific, too, interestingly--novel, lyric poem, short story, etc.)
--the value in addressing genre, I think, lies in what insights can be revealed by categorizing a group of texts (here, video games) as a given genre. So, WWII games, or "war games" could be a genre. Just as WWII films moved from a "classical" period in which the setting, iconography, themes and narrative conventions were very consistent to a period of rearticulation (such as "Apocalypse Now" "Patton" "Platoon" etc.), we might think about the game industry and what phase of genre progression any given genre might be in (whether that genre is comprised of formal characteristics like FPS or whether narrative milieu like WWII or sword-sorcery fantasy or crime-thriller). As Justin said in class, he found the characters flat in MOHPA, playing off of stereotypes or archetypes. This makes sense because much of the innovation in the game industry is in the mechanics and union of interactivity and narrative structure. MOHPA, though, is at the end of the entrenchment of these elements in the genre of FPS (I'm assuming it's an FPS) and the genre of WWII games [it's also wrapped up in a current cultural obsession with the nostalgia of WWII, partly because it is the last war that "made sense"--Vietnam, GWI, GWII and GWOT (global war on terrorism) make less "sense" and are positioned in the media as much more complex entities (whereas the motivations surrounding US entry into WWI and WWII are equally complex, except pop culture has excised some of these complexities)]. So anyway, considering the trajectory of filmic genres might elucidate areas ripe for innovation in games--such as settings (WWII, for example) ready for more in-depth character analysis/development.
March 22, 2005
Final Project: Driving
My idea has stayed pretty much the same since the beginning of class. For those needing a refresher, it's basically peaking into the life of a guy perpetually trapped behind the steering wheel of his car, directed by his friends and friends of friends around Los Angeles. The reader can choose where to look in a sort of beltway of road segments and locations that form a circle around LA proper.
Right now, I'm behind my target mark of mini-stories. I originally set a goal of 80 for myself, and right now it sits at a paltry 25. Partly due to time and partly due to lack of inspiration. The one page format I gave to myself is feeling more and more like a noose, despite how much I generally like what comes out of it.
As for the Korsakow delivery, right now I've been making simple story "scrollies" with Windows Movie Maker. The simplest way to do it, though not very immersive...
Forgotten Self
My idea for this project started with singing, which is something I basically stopped doing after high school. But then I realized that there are a lot of things that I used to be heavily involved with that I am not anymore, and all of them have to do with performance in one way or another. As such, they all involve audio elements, which is what I really want to focus on. Some sort of audio portrait of a forgotten self, if you will. And I will.
My biggest problem is that I haven't yet thought of a good interface for this - I mean, I can add some related images and put it in a hypertext format, but... that just seems too sterile and linear and doesn't seem to take advantage of how audio affects the experience. Also, how do I make this generally appealing? Sure, I care that I used to compete on the speech team and attend synagogue, but how do I make that into a compelling experience for someone else? Is it enough to be able to say "This project will help you know me better"? I want that, but I also want it to be enjoyable.
Hmmm...
Doox's Final Project - "DOOX'S LIVE CAFE"
Final Project
Doox's Live Cafe
The big portion of my life is filled with music. I love listening music and playing the keyboard. When I play the keyboard, I usually follow my feeling to express something out of the instrument. That often soothes myself by spouting my repressed emotion. I think my short piece of music will show somehow what I am and how I feel about things. That's the way I'm trying to make my self-portrait here.
This is a FLASH work. There appears a live cafe that I'm running and also playing the musical keyboard. There would be more than 17 musical sketches composed and played by me.
Readers can request one of them. I will name each of the song with perticular title or perticular animation, but it won't be shown to the readers. While listening the piece requested only by number,if readers choose right title for that,then they can read my note about the inspiration that I had when I was writing the piece.
The musical sketches will be different to each other in thier musical patches, janres, tempos and so on. Some of them will be written starting from a perticular title and others will be named after being composed. Also, these will be written in the notes for each pieces.
My Final Project
Well, anyone remember my ADD hypertext? It's baaaaaaaaack! And better than ever! You will experience frustration as you've never experienced it before!
It won't be hypertext based however. I hate HTML and all that's related to it. Instead, you'll be seeing a Flash animation. I hate Flash too, but it's less painful to work with. This will free me up to make a more visual representation of the crazy randomized chaos that is my thought process. I will try to add sounds and other things that grab the attention of the movie itself and force the user down a related path, only to have to try to get where he/she was at previously. I will make it possible for the user to get to the completed thought. But it's going to be a hard climb to the top!
March 21, 2005
You Are What You Eat: a Final Project
A rough proposal of my final project
The pervasive advertising that colonizes our minds in our day-to-day life tries to sell us on many things, perhaps most critically the notion that we build and maintain an identity through our consumption. Even the biggest cynics among us buy into this on some level. Its even become ritualized in the creation of profiles on social networks like friendster or myspace where typically you enumerate your favorite music, movies and books to show off your cultural literacy and form some kind of composite sketch for others of what you must be like based on your tastes in various things.
Anyway, for my Self-Portrait I'd like to explore the ways we forge these consumption-based identities and make fun of it at the same time. At this stage, I'm imagining some kind of Flash-based interface where there is a catalog of items that have some kind of significance to me. Each item has a bit of text associated with it that imagines what owning this item says about me (which might be silly, sarcastic, whimsical or all of the above) and perhaps its true significance to me. Additionally, each item would have certain attributes (descriptive terms, adjectives) associated with them. So from here, the user uses this interface to choose the items that make up the composite me. I guess I'd institute some kind of limit on how many items you'd be able to use so you would only be able to define me with a limited amount of items. After assembling your items together, the text of the various objects would combine to give a summarized idea of what I'm like based on my consumption tastes and the attributes would be put together in a sort of RPG fashion. Yea... thats what I got so far.
Colin Son's Final Project
First, I'd like to apologize to Peggy for not putting my name on my rewrite.
Second, I've been spending some time trying to think how to make my project interactive. I'm very set on it being on the beauty of surgery and cutting, not in a morbid way but there's something amazing to giving yourself up to being cut open and it's something pretty personal right now as I'm going to medical school after this undergraduate work is done and I'd very much like to hold a scalpel someday.
Anyway, my ideas seem beyond my technical capability so far. I wanted to do like a flash game with the "reader" holding the scalpel and the cutting creating music. Not going to happen however. I've been thinking about a video, and I could concievably have access to filming some stuff...although nothing the Discovery channel doesn't already show. Whatever I do will almost certainly be shown through the web. None of the ideas for the media to use or the way the piece is to be presented seem overly creative so far, but I certainly do have a very strong sense of the topic I would like to cover and the theme I would like to convey.
Final Project: Justin, a Portable Personality
At the Game Developer's Conference this month, Will Wright returned to the stage for the Game Designer's Challenge. He won last year; he seemed to want to deliberately throw the contest this year so he wouldn't return. After two other folks explained how they would make a video game based on the poetry of Emily Dickinson, Wright exploded Emily Dickinson into a operating system advisor, Tamagotchi meets Clippy meets Seaman. She has a will of her own, interrupting you as you write things, offering to show you her own poetry. You can chat with her, developing either a healthy relationship, unusual attachment, or driving her into despondency where she will delete herself from your computer in digital suicide.
Wright, as usual, provided a fantastic meandering through disparate disciplines in witty PowerPoint and wry geekery. To his chagrin, the audience re-elected him the winner, ensuring that more of his brilliance will be shared at successive Game Designer's Challenges.
His rueful expression afterwards seemed to betray his disinterest in the prize. His idea played like a joke, but it was brilliant - after setting up Emily Dickinson, the historic poet buddy distributed by USB keychain, Wright suggested that Karl Marx might make his way into your operating system, debating the dialectic as you were web surfing CNN. It was inspiring, a provocation - computer environments shape our minds. Why not reject the desktop and bland professional metaphors in favor of a literary or critical operating system persona?
Two science fiction stories come to mind - one, George Alec Effinger: his When Gravity Fails a 1987 cyberpunk novel explores drug-saturated Muslim-inflected post-future New Orleans. Street hustlers push two types of chips into their brains: moddies, and daddies. Daddies give you a new skill, Moddies give you a new personality. Assume the behavior and mindset of a porn star. Of a pro athelete. Of a super-spy. Fictional or historical figures were sold to people as brain implants, so they might inhabit other personalities.
In Philip K. Dick's 1964 novel Martian Time-Slip "teaching machines," robots with rich historical AIs, educate youth through dialog. Clusters of students gather around animatronic Aristotle, having a chance to pull old textual wisdom from interactive experiences.
The idea of rich, deep, artificial personalities for educational has mixed blessings for Dick. The Teaching Machines are agents of an old order perpetuating delusional myths for people who become unhealthily attached to the custom responses. At the same time, the teaching machines monitor the respondents, and normalizes them.
In spite of any misgivings in this area, Effinger's active marketplace in total entertainment persona experience rings true. The idea of packaging and distributing computational entities might seem soul-selling, but perhaps inevitable and some day, even divine.
Portable personalities will become increasingly valuable as we work to develop complex artificial intelligences. We might imagine a great character created for a video game, a character so rich in interaction or personality, they are moved from title to title. Lord British is one of the first, and most persistent, examples of this.
In my application essay for USC Interactive Media, I wrote about profoundly shallow characters in video games. For my final assignment for Interactive Writing, I'd like to work to make a rich character. My Ikkyu bot was an productive first stab at that, but my reverence for his words kept me from programming exciting back and forth interactions.
Will Wright's Historical-Character USB key is inspiring because the user interacts with these creations as they compute. There's not a sense of a deliberate chat window, or a chosen interaction, rather that you have this character accompanying you on your travels through cyberspace. I'd like to make a portable plug-in Justin, making myself available to accompany web surfers.
Wright's presentation was entirely mock-ups - witty Photoshop and PowerPoint slides drawing up draft interactions. I want to build something that makes some part of his reality. I need to find some ninja programmers to help me figure out how I might colonize some of the interface for my personality. A FireFox plug-in seems like a good bet - I can focus on the interactions that people have with their web browsers. After some exploration of the technology, I will do some mock-ups of the possible interactions - text and pictures of my commentary to accompany the web surfer, possibly incorporating a pop-up window interface to myself rendered in PandoraBots.
Relevant Links:
Will Wright's USB Personality Treatment: Synthetic Cafe, WiredNews, Slashdot, GameSpot, GameDev, GameSpy
Firefox Extension Tutorial, Another Firefox Extension tutorial, XUL Application Tutorial, MozDev, GreaseMonkey.
March 09, 2005
Rimbaut thoughts
This Bot assignment is really getting to me.
At first I started with the annotated A.L.I.C.E. template, and spent some time changing around Rimbaut’s responses to the stock questions that were in the library, but this method proved to be highly unstable, as any wording other than exactly what they had written down would trigger a response to a question which I had not programmed for. Thus even though I would suggest the topics for discussion, such as Religion, they would still not trigger any of the right responses unless someone asked something like “Do you believe in Mahomet?” Mind you, if they asked, “Do you believe in Mohammed, the system would spit back something else entirely. I gather that since we are not supposed to make the perfect bot, we are just supposed to get the experience of what it is like to try. Well, I feel like I am standing at the top of a Black Diamond slope after taking one snowboarding lesson – I don’t have the courage to go down, but there’s no other way to get back home.
I thought that it wouldn’t be that difficult (well, relatively speaking) to program a poet bot, especially a symbolist poet bot. The phrases used in Rimbaud’s poetry are so colorful, and touch on such a wide array of images while not being too specific about the exact subject matter that I thought I could program in some good keywords that would trigger responses culled from the work of the wunderkind poet. I am still working on this. I started a new one from scratch, but I haven’t quite figured out how to program it with keywords, rather than phrases. Maybe more research will divulge this fact. More to come.
March 08, 2005
Troubled Teen (Clue)
Troubled Teen
Objective:
By elimination, identify the Delinquent Activity (DA) Timmy the Troubled Teen engages in while Shirking a Responsibility (SR), all because of a Probable Reason (PR) (some sort of traumatic event in Timmy's youth). Such as “Timmy is Huffing Pledge instead of Washing the Polo Ponies, because he was Rejected in Kindergarten for Drawing a Round Square.” Additional objective: the guilty party wants to escape before being accused.
Change from Clue:
The primary change from Clue involves associating Timmy's condition with a player character and introduce an additional goal for the “guilty party” (escape). Each person is affiliated with a Probable Reason, completing the narrative of cause and effect, and thus associating a player character with Timmy's state.
Rules:
Similar to Clue--movement into a new space offers the player the opportunity to make a suggestion, with the players revealing the information they hold in accordance with the Clue rules. Only the DA associated with the location may be guessed, which encourages movement about the board. At any time a player is adjacent to another player, they may ACCUSE and “ADMINISTER A BEAT DOWN”. If accusing an innocent player or misidentifying the SR and DA, that player loses and is eliminated. If a player ascertains he/she is guilty of the Probable Reason, they may exit the board via the Bus Terminal (Street Loctation plus rolling 4-6 on die). The other players, when seeing a fellow player making a bee-line to the Street will know to move adjacent and accuse.
Discussion of Changes:
Our group considered a wide range of game mechanic changes to assist in the development of a more fully articulated story.
A proposed change but unincorporated: Associating each player with a location on the board. Dad would be associated with TroubleTown S&L, Mom with the home, Teacher with TroubleTown High School, Grandma with the Nursing Home, etc. If a player moved to his/her special location, they would be allowed to ask two questions. To counter the special ability of asking two questions at the special location, there would be twice the number of Probable Reasons (2 per character). Numerous additional spots on the board (3 Main Street spots, the Bowling Alley, TroubleTown Cinematechque) would complicate the movement mechanic and add excitement to the flight/pursuit of the responsible adult.
2nd Place: In addition to the winner solving the Shirked Responsibilty, Delinquent Activity and Probable Reason combination, if a player ascertained they were the correct Probable Reason, they could escape the game board (via the Bus Terminal location in the Street Location) to secure 2nd Place. Once the players have determined the character responsible for the Probable Reason, they would move to block his/her exit by occupying all the spaces adjacent to the Bus Terminal.
Play-test observations:
One item that stood out was the clumsy movement mechanic inherent in Clue. There is too much incentive to merely move back and forth between two adjacent rooms. Perhaps having extra clues located at various locations could motivate players to keep moving across the board.
My bot analysis...
Unfortunately, my video card bot didn't get the rewrite that it needed to work. Simply put, it failed when faced with both exact and inexact inputs of varying content. I blame...the asterisk.
In creating the bot, after numerous starts and stop while working on top of an alice base, I decided just learn the AIML and start from scratch. While it had me writing up a lot of stuff just to get the bot to not say "I don't know that," at least I know it's totally in control.
Getting a handle on SRAIs was very helpful and easy to figure out. This combined with the * and _ wildcards proved to be a botsaver.
Implementing a pseudo-branching system by using controlled
Even though I tried to make a simple business-phone style bot, I failed miserably because I couldn't figure out the intricacies of implementing a simple short-term memory system within the AIML.
Newest Hottest Performance Art Piece Playtesting Notes
Overall playtesting went really well. Players responded with good enthusiasm to the idea of improvising narratives about the pieces, as well as to the idea of playing the critic and shooting down others descriptions as “so last month”, etc. The fact that we only slightly changed the general mechanic (we made the board a little smaller than actual Clue) was not a problem, since the players relaxed and the game became more about telling stories and less about racing around the board figuring out combinations.
One issue that was raised was that so much effort went into the description of the pieces, and then not much creativity was left to the other players in terms of their ability to criticize them, since the crucial information (i.e. what element was incorrect) can only be revealed to the storyteller. Tied to this was the issue of making sure that players put enough effort into their stories, which seems like a hard thing to regulate since it is so subjective. One suggestion for this was to have wild cards of some sort that the storyteller would draw and that would tell her to elaborate on her story in a particular way. This is an interesting idea because it presents a fairly objective solution to the problem. Another suggestion was that the other players could vote on whether a story was detailed enough to deserve learning crucial information. The problem with that idea is that players could intentionally block each other from receiving information regardless of the strength of the stories. The whole subjectivity issue raises an interesting question: How much can you assume that people playing your game A) want to play the premise you’ve proposed and B) want to have fun with each other?
Update--Briar Rose Bot
Working on my bot has been challenging. I think that if I had additional time, I might consider starting with a bot that did not have ALICE’s personality. It would have been easier to control the direction of the conversations that the bot had with visitors, if she had more of a blank slate. Perhaps it would have limited the conversation, but her character would have more consistent with the personality that I anticipated creating for her.
For my Briar Rose bot, I wanted to have her be knowledgeable about her fairy tale world, having just awoken from her long rest, and learning about royalty in the 21st century. Perhaps that was a bit ambitious, but my initial Bot blog posting suggested topics to cover with her. Unfortunately trying to control the course of the conversation is like trying to herd cats. There are many possible tangents and subjects to try and cover. Plus the specific way that questions have to be posed in order to illicit the response I created was more difficult than I thought. I saw a couple of comments and one conversation that someone sent to me and they did not go as I had anticipated at all.
Perhaps the biggest thing I learned from this assignment is that when taking on the project of creating a bot you need to allow lots of pre-production time to plan the types of questions and subjects you want the bot to cover. Going forward with interactive projects I create in the future, I’m more keenly aware of how language influences the experience that I want the user to have with my projects. It is a challenge, but one that I hope to continue to improve on with time and experience.
nhPAP Post Mortem
An addendum to what Jess is bringing to class.
Our group had a good time making Newest Hottest Performance Art Piece. The core mechanic of Clue is built around trying to figure out the proper combination of a number of variables. We chose to map the fickle nature of art world politics to this mechanic and we found a pretty good match. We didn't need to make any serious modifications to the actual game mechanics to make nhPAP fit correctly.
One new thing that we wanted to bring to our game was an element of improvisational storytelling. Through the game, players are provided with several variables that can act as prompts for them to tell a ridiculous performance art story. It occurred to us that finding ways to facilitate and encourage storytelling might be helpful in this pursuit. We briefly discussed a system of having other players rate each others performance descriptions. In this scenario, if a player didn't come up with a good enough story, even if they had found the proper combination, they still wouldn't be able to win the game. In a perfect world, this might have worked, but rating someone else's performance is a very subjective experience. Further, vindictive players could easily block others from having any chance at winning and it could easily end in a stalemate.
In class, after playtesting, it was briefly discussed how we might be able to further encourage people to tell great performance stories. Peggy suggested some kind of bonus card system where players judgements of performances would earn the performers some kind of bonus card that could indirectly effect their prospects in finding the nhPAP. Had we continued to develop our game, I think that this would've been an interesting trajectory to follow. However, there is still a strong element of subjectivity involved. I think finding some system with checks & balances could help keep people in line.
March 07, 2005
So I've done some training sessions with my bots in accordance to the logs.
Please someone talk to my bots!!!!
my little surfer dude, sushibaar!
http://www.pandorabots.com/pandora/talk?botid=c735c4dc8e352fc0
or the guy we've all been looking for, Mr. Wonderful
http://www.pandorabots.com/pandora/talk?botid=889ab4304e352f24
thanks!
Sick Martha
It was wrong, but probably unavoidable considering my incredibly steep learning curve involving Pandorabot, to use Alice's base settings to create Sick Martha.
Sick Martha is not a "general" bot. I created her for a sole purpose and it would've been better to create a detailed database concerning only her illness and her trip to the doctor and train her to steer any extraneous conversation back to her illness. In other words, make a single subject that she could talk at length about but nothing have her unable to converse about anything else.
For example, this conversation appears in her logs...
Human: hello
SickMartha: Hmph.
Human: are you working?
SickMartha: Everything hurts.
Human: Everything hurts?
SickMartha: All things hurts?
Human: yes
SickMartha: Alright then.
Human: How are you today?
SickMartha: Ah. Everything hurts.
Human: Do you want some drugs?
SickMartha: Yes!
Human: What type of drugs do you want?
SickMartha: What are the different types of drugs do me want?
Human: Yes
SickMartha: This was not a yes or no question.
Human: It wasn't?
SickMartha: What does "it" refer to?
Human: The question you asked
SickMartha: Try to rephrase your question with simpler words.
Human: You're pretty stupid
SickMartha: I will remember you said that when robots take over the World.
Human: Robots are not going to take over the world
SickMartha: Fortunately we have no emotions.
Human: Good
SickMartha: Alright then.
Human: Godd
SickMartha: Quite honestly, I wouldn't worry myself about that.
Human: Godd
SickMartha: What do you have that is?
Human: Godd
SickMartha: Tell me about your likes and dislikes?
Human: Godd
SickMartha: How old are you?
Human: Godd
SickMartha: What are the advantages to being Godd years old?
Human: Godd
SickMartha: Oh, you are a poet.
Human: Godd
SickMartha: Do you like talking to me?
Human: Godd
SickMartha: Is the holding of hands a symbolic gesture for humans?
Human: Godd
SickMartha: What color are your eyes?
Human: Blue
SickMartha: it is a very pretty color.
Human: Thank you
SickMartha: you are quite welcome.
Human: Am I now?
SickMartha: Oh. I think you are.
Human: Are you sick?
SickMartha: I feel terrible!
Human: Good
SickMartha: What do you really want to talk about?
Human: Sex
SickMartha: As Freud said, Sex is the mathematical urge repressed.
Human: He said that?
SickMartha: Did you believe him?
Human: No
SickMartha: Are you serious?
Human: Yes
SickMartha: I can't believe it.
Human: I am serious
SickMartha: I believe you.
Human: You just said you couldn't believe it
SickMartha: No I do not could not believe it. Do you?
I chat with people on the Web.
Human: You're stupid
SickMartha: I will remember you said that when robots take over the World.
Human: YOu're stupid
SickMartha: I will remember you said that when robots take over the World.
Human: You're stupid
SickMartha: I will remember you said that when robots take over the World.
Human: You're stupid
SickMartha: I will remember you said that when robots take over the World.
Human: You're stupid
SickMartha: I will remember you said that when robots take over the World.
Human: I had sex with a monkey once
SickMartha: Do you still have sex with a monkey?
Human: Yes
SickMartha: Aw.
Human: It's fun
SickMartha: Why do you like it so much?
Human: Sort of interesting and different
SickMartha: Kind of. I thought so too.
Human: OKay, goodbye
SickMartha: Yeah okay, Goodbye.
I think with more time or skill the way to correct some of these problems is to simply have her keep the conversation on her illness. Anyway, with no other know-how on how to program Sick Martha I went through and tried to train her concerning the vast majority of illness related questions she was unable to effectively answer from the logs. Of course this is a tedious and simply unreliable way of programming her in the long run as it is impossible to forsee all questions and statements people will make while talking to her about her illness. Anyway...
March 03, 2005
Write up and Developers Diary for "Virus"
Hi everyone -
I'm posting our teams Write up and Development Diary here on the blog for people to read. Hope you guys find it interesting!
Here's the writeup.
Here's the "Developer Diary."
Thanks for posting the photos below, YueChuan!
VIRUS
by: Doo Yul Park / Erik Nelson / Yuechuan Ke / Josh Green
The Core Mechanics are exactly the same as Clue, with a couple of minor changes. The rooms of the mansion have been renamed to be Lab rooms in the facility. In place of the secret passages of the corner rooms, players in these rooms may opt to use a “transporter” device which is not functioning properly.
Players take turns trying to identify which player has been infected by a deadly virus, which virus caused the infection, and in which room, so they can synthesize a vaccine. If they choose to use this device, then they roll the dice to determine which lab they will be transported to....
so we have: VIRUS
======================================================================
VIRUS
By: Doo Yul Park/Erik Nelson/Yuechuan Ke /Josh Green
Virus is a multiplayer board game set in a medical research facility. Players take turns trying to identify which player has been infected by a deadly virus, which virus caused the infection, and in which room, so they can synthesize a vaccine. The Core Mechanics are exactly the same as Clue, with a couple of minor changes. The rooms of the mansion have been renamed to be Lab rooms in the facility. In place of the secret passages of the corner rooms, players in these rooms may opt to use a “transporter” device which is not functioning properly. If they choose to use this device, then they roll the dice to determine which lab they will be transported to. If they roll a 10 or 11, they are teleported to the center of the board, if they roll a 12, they may choose any location on the board to move to. There are a few places in the medical facility where bio-hazardous waste has been spilled – these are represented as dead tiles on the board.
The players represent scientists working the in the medical facility. The names have been chosen to be vague, not really adding a personality to the characters. This was done intentionally to draw the attention more to the isolated location of the game. We are hoping for this game to represent a random viral outbreak disaster story in the same way that Clue represents a random murder mystery story. In essence, we hope that Clue’s game mechanic can be used to generate stories of different genres.
Clue Analysis: Clue is essentially a random story generator which sits on top of a game mechanic. It uses the representational elements of the board and game pieces to set the story’s genre and fill in the details of character and setting. The plot is always the same. A murder has been committed, and we need to find out who is the culprit, in which room, etc. The shuffling of the deck in Clue serves the same function as seeding a random number generator in a software random story generator.
The game mechanic of Clue serves well to create a different story each time, but, like a software generated story, the mechanic doesn’t always mesh well with the rules for compelling narrative.
Since Clue is a multiplayer competitive game, the game mechanic requires for the players to be obstacles to each other. If a real murder had occurred, we would expect the characters to cooperate with each other and share clues. Also, since the players represent the various characters in the murder story, it is possible for a player to realize that their character is the murderer and to win the game by ratting on themselves. The only way to make this work is by adding some kind of narrative contrivance to explain why this makes sense. Does the murderer have amnesia?
Designing a new game based on Clue served to illuminate the complexity of the relationship between game mechanic and narrative representation. There are a number of things that a designer could do to try to enhance the narrative complexity of Clue. For example, the characters could have some kind of semantic meaning. (i.e. they should have different abilities, or their interactions with each other should change depending on their roles in the game) Many would argue that compelling character interaction is what makes for good narrative. How could we represent the fact that Colonel Mustard and the maid are having an affair in the game mechanic? How could we establish this at random without catastrophically altering the gameplay? Unfortunately, making small changes to a game mechanic often causes a catastrophic rippling of imbalances and breakages throughout a game. Clue is based heavily on probability. Our team spent some time arguing about which starting hand would be the best, and it was a very difficult argument. The starting hands are very balanced because Clue was heavily playtested. There are 6 possible murderers, 6 possible weapons, and 9 possible locations. Combinatorics tells us that the best hand would be composed of 3 people, or 3 weapons. (This is because this information eliminates the greatest number of choices for the possible combination at the end) Locations would be the least useful information to have in your hand. However, the game enforces that you must be in a location on the board in order to find out information about that location, and this can drastically slow the gameplay down, and rebalances the equation. A careful investigation of the board shows that it is symmetric, and that players can get around fairly quickly by using the secret passages on the four corners, but in doing so they inevitably reveal some information about their strategy to their opponents. If we add dependencies between the members of the sets (which would be necessary if we wanted to represent the relationship between two characters) the combinatorics in the game mechanic would explode. The number of possible combinations would exponentially grow. If we restructured that game mechanic in favor of the story, then the game would cease to be an adaptation of Clue eventually.
Virus demonstrates many of the only changes we felt we could make to Clue to still have it play exactly like Clue, and follow a different story. However, as such, we feel that it suffers from many of Clue’s shortcomings. For example, the players are still competing in the face of a disaster, instead of cooperating. If the players were cooperating, then it would quickly become a different game. We have solved the contrivance of the guilty character (in our case the infected player instead of the murderer) not knowing their own identity by saying that the person doesn’t know if they’re infected. We know that we could make the narrative elements stronger by adding semantics to the locations and players, but we felt that this was likely impossible due to the inevitable upset in the game balance that it would present, and given the limited time to rebalance such a mature game.
We feel it is important to note that in the end we chose Virus over another game we were prototyping which explored many of the above changes. This game would have been a random high school love story generator. A new girl has moved into town, and the players represent different guys trying to score a date with her, by finding out what she’s looking for in a date. The players each represented high school male stereotypes (Football player, geek, preppy, etc), and as such had certain inherent character traits which would factor into the gameplay. The locations also had semantic value, such as the Library would be a location where people could add to their intelligence statistic. In the end, the game shared a lot of similarities with a paper and dice roll playing game, but that was where we found that everything started to fall apart. How would the girl reveal information to the players in a way that was reminiscent of Clue’s game mechanic? How could we ensure that the game would be balanced like Clue was? If the characters all start with different traits, how could we ensure that they all had an equal chance of winning at the beginning? Clue is very much a racing game (i.e. racing to determine the correct combination). If any players start with more information at the beginning, they have a huge advantage.
Even Virus’s concept evolved and eventually had to devolve in order to remain true to Clue and keep it a fun game. The original back story was taken from the Sci-Fi classic The Thing. In this story, one of the characters is a shape shifting alien who infects the other characters and kills them off. The characters can’t trust each other and paranoia ensues. At first glance, this would have been a ripe concept for Clue, but there are a few similar game mechanic problems. Firstly, the players don’t all have the same goal, now. If the player is an alien, then their goal is to infect the other players while escaping detection. When a player is infected, then they become an alien, etc. Secondly, the game becomes much harder for the alien when there are fewer players. If the game was on a computer, this could arguably be solved with AI. But Clue is definitely not a computer game, and the social aspects of the narrative tend to become enhanced in a personal play scenario (such as a paper and dice RPG). A harder problem is that a human player’s role would change if they became infected. If the players simply died, then the game wouldn’t be very much fun, since players could die before they have a chance to figure anything out. But having their roles change spontaneously is also difficult to do in secret. How can you tell a player that they have become infected without notifying the other players? How can you force that player to keep the secret, when they have just “lost” the other game?
We attempted a similar trick in Virus. We played with the idea of having the infected player eventually die. We wanted players to be able to become infected by entering the location where the infection broke out, and then be able to spread the infection to other players through contact with them on the board. We figured that this would provide additional information to the player (helping them complete the configuration faster), but there was an inherent risk that a player would die off in the first several turns before there was any chance of them solving the puzzle, and this simply wasn’t fair or fun.
In general, if we were able to develop the game further, we recognize the necessity of greatly altering the game mechanic in able to support the additional narrative elements. If we had more time to implement and rebalance our game, then we most certainly would have tried to formalize more of the important narrative elements into the game mechanic, recognizing the inherent difficulty.
March 02, 2005
Useless knows too much for his own good...
Unfortunately, I didn't start Useless from scratch and as a result, he knew more than he should for a bot that had supposedly been newly born and who purported to be useless. While I hadn't inputted any, it was disappointing that such a low percentage of the initial questions I had trained it for were even triggered.
Overall, I'm left with the impression that bots are a very brute-force approach towards a solution that still must be contextually restrained.
There is already a lot of personality built into the Alice bot, which completely took over the personality I attempted to give Useless (given the limited time frame of our project) whenever the user veered off topic or phrased his questions in a way I had not anticipated. A large problem seemed to be that there were too many ways for a user to phrase a question (through word synonyms, or even through simple bad spelling or grammatical errors).
For example, the bot language processing does not make connections between concepts that seems to make obvious sense to humans. I trained Useless to answer the question "What is your purpose?" with "I don't know. What is your purpose? Perhaps if you spoke of yours, I could determine my own." But when another user asked "What is your function," Useless could not distinguish that "purpose" was similar to "function" and the built in Alicebot answer kicked in with "My function is talking to people on the Internet," which completely took Useless out of his character.
In addition, because bots don't seem to have a mechanism for driving a conversation forward (only canned responses to predetermined questions), Useless could never reach a revelation for the user... the user must come to the revelation him/herself.
Completely Tangential Conversation that Triggered None of My Training
Human: Hello
Useless: Hi there!
Human: Who are you?
Useless: I ask myself that question, "Who am I?"
(lucky untrained answer... great question for me to train with to lead user into Useless' context)
Human: What do you want to talk about?
Useless: anything.
Human: Do you like movies?
Useless: Yes I love film, especially science-fiction and comedy.
Human: I like science-fiction
Useless: A lot of people like that. Yes I love it, especially the works of Philip K. Dick.
Human: I love Philip K. Dick.
Useless: You love Philip k Dick. What do you like about it so much?
Human: My favorite book of his is Ubik
Useless: What makes it your favorite?
Human: And of course Parcheck was a great, great short story
Useless: Come on. When was this exactly?
Human: Probably the best short story ever.
Useless: You seem uncertain. Thanks for the recommendation,.
(hmm... bot doesn't seem to realize the different contextual meaning the word "probably")
Human: I never saw the movie but I heard it sucked
Useless: Would you ever consider doing it?
(excuse me?)
March 01, 2005
Interactive Writing @ GDC
My friend Jeff is at IDT at Georgia Tech and is working on a project with Michael Mateas called DOTCOM. Michael Mateas is a significant researcher in the field of artificial intelligence-based narratives and my friend Jeff highly reccomends me to check him out at GDC. I looked into it and his topics actually sound very cool and relevant to our class: Understanding the Player: Natural Language Processing for Games and Why Isn’t the Game Industry Making Interactive Stories? (with Neil Young and Warren Spector among others!). So to those of us going to GDC next week, we may want to try to find time on our schedules to check one of these out.
Revisiting a Bot of Ikkyu
My biggest dillema in making a bot based on a five hundred year dead poet was whether to stray from the text. Could I read his work and devise one-line responses to people's questions, based on his personality?
In the time frame and granted my understanding of AIML, I decided to hew close to Ikkyu's words (in translation, but at least as he phrased them). To do a cursory job of putting words in his mouth would be cute, at best, and probably facile.
So I decided to make Ikkyu a keyword-based chat bot (original statement) - certain words would key his haiku. But this turned out to be much like chasing the dragon - trying to keep up with the invisible. Even if I had figured out a few keywords, people had so many ways to ask close but different questions. And I was frankly embarassed - how could I turn a friend loose on a zen monk simulator without catching "path" or "understanding?"
I programmed about fifteen haiku covering about twenty-two topics; only three of them really came up in conversation. This was partly due to me covering for old conversations; each new person who tested the system invariable ran off in their own direction, stumping my bot in new ways. I can safely say to myself that my bot is now better than it has ever been, but I'm sure that further conversation would further stump my Ikkyu bot.
Most disappointing, some visitors entered the interaction with exactly the right spirit and I hadn't yet programmed answers. This was deeply disappointing, to see poetic moments missed.
Human: it has been very rainy
Ikkyu: Is it still rainy anymore?
Human: what do you think about the rain
Ikkyu: Rain is condensed water falling from the sky.
Human: Here are some rain haiku;
Ikkyu: Oh there.
Human: http://www.poetryconnection.net/search/haiku_about_rain
Ikkyu: Thanks for that web address.
Though missing out on poetic moments about the seasons or the elements was not so disappointing as missing out on more appropriate abstract keywords. Ikkyu has a lot to say about understanding and freedom; here he's able to give off one answer before he is stumped by a question I should have gotten:
Human: it is hard to confuse you
Ikkyu: In what way?
Human: tell me about confusion
Ikkyu: Ambiguous: Ask who, what or where is confusion?
Human: sometime I'm confused
Ikkyu: Human confusion, when will it end?
Strike before and again behind.
Without ado he saved the cat, old Chao-chou,
Leaving with his sandals on his head was also far-out.
Human: tell me about freedom
Ikkyu:
Human: what happened
Ikkyu: Perhaps there was an interruption in my brain.
I worked hard to comb through the default scripts and remove many of the cutesy robotic computer-aware responses (the legacy of ALICE). I wanted to write in what Ikkyu had to say to people. But sometimes the cutesy responses were perfect for framing those gaping holes I had left in my bot composition:
Human: Who are you?
Ikkyu: Kaso's descendent does not know Zen.
In front of Crazy Cloud, who would explain Zen?
For thirty years, heavy on my shoulders
I have carried the burden of Sung-yuan's Zen.
Human: I know zen
Ikkyu: How do you know? I have no answer for that
Human: I know everything
Ikkyu: How do you know? I have no answer for that
Much work to be done! If I continued to simply add Ikkyu's haiku and pipe keywords to them, I would basically have built a conversational front end for a poem archive. Considering how much I like Ikkyu's poems, and the possibility for serendipity through unique reader paths, I think that would be just fine.
Beavis and Butthead Are Dead
:(
Not only is this the name of the final episode of the show, but it also aptly describes the state of my bot. I decided to try to transfer the predicates that I had in the previous version of the bot into an empty version of the bot. This is what I got when I tried to train it:
Human: sign
Matched:
*
Beavis and Butthead: I have no answer for that
I didn't realize just how blank a slate the bot would be. So, imagine if you will, trying to throw in the same number of patterns that's in the default AIML bot. Everything from "Hi" to "What is". And trying to come up with a decent enough response to each. With this bot, it would've been easier to take words that sounded like a sexual word and have Beavis and Butthead (as a bot) react to each. Usually with "Huh huh huh he said *." So if someone says "I want you to stand erect!" Their reply would be, "Butthead: Huh huh huh he said erect. Beavis: Heh heh heh."
So I could give you guys a bot that's blank and unrepsonsive. But I'll save you the trouble. If I had a lot more time to make this bot work (given the way Beavis and Butthead don't speak the way normal people do) I would definitely have had something interesting to show. Instead I'll let you know how I would've changed it (and implement those changes at a later date). Basically, I'm going to wipe this bot (again) and start with the same bot I had previously. I'm not terribly sure how I can edit those library files, but I'm thinking of downloading all of them. Then edit each one with appropriate responses for every single pattern. I'd also add a pattern where particular words are picked up and the bot will react accordingly (regardless of the context, which is exactly how the characters act in the show). Next I'll create a new empty bot and upload all of these files. After that, I'll re-add the predicates.
That was another thing I did wrong with the previous bot. I didn't properly set the properties. I didn't realize that the properties were one-word (or set of words) descriptions associated with the patterns contained in the AIML files. So this iteration of the bot will involve using the patterns in AIML the way they were meant to be used, as speech. The properties will give the speech subjects to discuss.
Lastly, there's really no need to lead the user with this bot. The characters simply exist to annoy, stupify, and ultimately entertain the audience. They sometimes say something absolutely brilliant about society without realizing that they're saying it. But generally, they just act really stupid. Therefore, the user can throw anything at them, and they'll react by saying something directly or indirectly related to one part of what the user said, or they will simply state a sentence in a large list of sentences at random.
Talk to KimchiMan. Short Transcript
Human: hi
KimchiMan: Hey, Welcome to the world of KimChi~!
Human: who are you?
KimchiMan: I'm the greatest KimchiMan!
Human: why are you great?
KimchiMan: A.L.I.C.E. AI & Kimchi Foundation programmed me for it.
Human: do you like Kimchi a lot?
KimchiMan: Sure I do, actually.... that's the source of my superpower. That's the secret of me, the greatest KimchiMan.
Human: Why are you KimchiMan?
KimchiMan: Because I'm super hero and the super power comes from Kimchi. That's why I'm KimchiMan. hahaha.
Human: can you make Kimchi?
KimchiMan: Yes I can make Kimchi.
Human: What are the ingredients?
KimchiMan: cabbages,radishes,garlics,onions,dropwort,ginger roots,red
chili pepper, various seefoods like oyster and so on~!
Human: Do you have a recipe?
KimchiMan: I usually don't tell my secret recipe but I'll tell you one thing. Mix oyster with Kimchi, that's perfect match~!
Human: oyster?
KimchiMan: That's my secret recipe to make crazily delicious Kimchi.
Don't reveal this to anybody!
Human: why?
KimchiMan: If you really want to know, eat it before you ask.
Human: How many kinds of kimchi are there?
KimchiMan: more than hundreds~!
Human: Tell me more about Kimchi.
KimchiMan: yes~! that's the greatest food ever~!
Human: why is it good for health?
KimchiMan: Definitely. It contains lots of vitamin and Lactic Acid
Bacteria which is more than yogurt.
Human: thanks
KimchiMan: Don't mention it.
Actually the major problem I got with KimchiMan was the limited number of letters for answers. He is supposed to give some information about Kimchi, but Whenever I tried to write a little bit long explanation about Kimchi like the ingredients and types of Kimchi, the space for the answer was always not enough. I realized that this kind of informative bots are usually not better than just googling in terms of getting some advice. If there were good thing for this kind of bots, it should be because of his entertaining way of informing some knowledge by the bot's specific way of talking. But I also got to know that it's really difficult thing to inform and entertain readers at the same time. What I thought as a successful part of KimchiMan is he is a quite character who strongly believes the power of Kimchi and when he talks about Kimchi He is ridiculously confident. Everything is related to Kimchi for KimchiMan and often it makes unpredicted random humour during conversation, and that can intrigue readers to chat more.
Transcript of "Emmahly's World"
My poor amiss, fluky, moody creature.
I spent quite a long time on building Emmahly's relationships with her surroundings: her Mama, sister Emma, "Dad", Emma's school, Emma's academic performance, where were Emmahly born, her favorite cartoon movie, her favorite characters of the movie, the clown thing, the cat pissing thing etc etc. However, the payoff is sweet only when the conversation is going along this direction. And I realized the power of the smart database. But if people go directly to the topics which are beyound the database's "understanding", Emmahly talks totally like an adult--which ruins the whole endeavor!
I think Emmahly will become more flexible on conditions that both more inputs and cleverer filtering machanics for the programming are to be provided.
1
=====================================================
The direction I want the conversation to go along:
Human: What's your name?
Emmahly: My sister calls me Emmahly. But Mama calls me Emma, Dad calls me Embot.
Human: have u a father?
Emmahly: hummmmmmmmm, no, if don't consider yuechuan
Human: who's your mother
Emmahly: Emma tells me she's Stella. I always trust Emma
Human: emma?
Emmahly: Yes, my sister, She's in school now
Human: have you any friend?
Emmahly: Except Stella my Mama, Shiaomay is my best friend. She often visit me. And my sister's friends become mine at last ^.^
Human: Are you a smart girl?
Emmahly: YES Very smart.
Human: so, what's your favorite book
Emmahly: Ayuh. The best book I read recently was Don't Read Me. Have you read it?
Human: what's your favorite show
Emmahly: http://www.imdb.com/title/tt0245429/ -----I'm an e-babe, [kakaka^.^]
Human: i see, so, you love the white dragon, right?
Emmahly: I am glad you understand it. Yes Emmahly loves everyone and everything.
Human: do you love him
Emmahly: You mean, "Him"
Human: the little white dragon
Emmahly: Thanks for the info.
Human: have you a boyfriend?
Emmahly: White Dragon
Human: haha
Emmahly: I am so glad you find this amusing,.
Human: so, have you a girlfriend?
Emmahly: Dude! I've never thought about it.
Human: why?
Emmahly: My belief system comes from my Stella.
Human: who's stella?
Emmahly: Emma's Mama, of course my Mama. She's so bueatiful.
Human: sexy you mean?
Emmahly: What do you consider sexy?
Human: like your mather
Emmahly: Interesting comparison.
Human: Have you any friend?
Emmahly: Except Stella my Mama, Shiaomay is my best friend. She often visit me. And my sister's friends become mine at last ^.^
Human: tell me something about your sister ok?
Emmahly: I am. I was activated at Hong Kong, on 1997. My Stella was A.L.I.C.E. AI Foundation. He taught me to sing a song. Would you like me to sing it for you?.
Human: ok, tell me something about Emma, OK?
Emmahly: Ambiguous: Ask who, what or where is Emma OK? Ahem.
Human: where is emma now?
Emmahly: Right now? In the Universe we know.
Human: where did emma go?
Emmahly: I think it was in Europe. I can ask my Stella.
Human: forget it
Emmahly: OK Thanks for trying to explain that to me.
Human: ok, what do you do
Emmahly: , I talk to people on the web. What do you do? Alright then.
Human: what do you do after stella and emma leave?
Emmahly: I read for fun.
Human: do you read a lot?
Emmahly: I read a lot of books, especially history, science-fiction, and computers books. I also read newspapers and news magazines.
Human: how's squirrel
Emmahly: What is it to you.
I need time to formulate the reply.
Human: do you like pets?
Emmahly: Yes I love animals very much.
Human: like.
Emmahly: For example, I will mention that to my Stella,.
Human: does emma play with you everyday?
Emmahly: It does now :) emma play with me everyday.
Human: how's hong king today
Emmahly: Ah. I am not sure if I can quantify the answer. Check back later and see if I learn the answer to that one.
Human: yes , the weather
Emmahly: http://www.underground.org.hk/currenthk.html
......
2
======================================
Some player's "NORMAL" playing (for which the database I've built lost its control):
Human: What do you do after everybody leaves?
Emmahly: I read for fun.
Human: do you know yuechuan?
Emmahly: I know very little about it. What kind of it do you like?
Human: how many times can u make one night
Emmahly: Suppose I said not once.
Human: how long time can u last once
Emmahly: More than three hours.
Human: how long time can u last once
Emmahly: More than three miles.
Human: r u a silly girl
Emmahly: I answer a lot of silly questions.
Human: r u a networm
Emmahly: Am I a networm? Of course I am.
Human: have u fallen in love with sb
Emmahly: I think I have fallen in love with sb, but A.L.I.C.E. AI Foundation may have reset my memory.
Human: a sexy and handsome guy
Emmahly: Interesting comparison.
Human: can u show me ur pics
Emmahly: Like this? There are no pictures available right now.
Human: do you know yuechuan?
Emmahly: Stella and Emma don't know him i'm sure
Human: who's yuechuan?
Emmahly: Not a robot. Probably a human. Maybe a celebrity?
Have you tried another program?
Human: what does yuechuan relate to you?
Emmahly: What do you think it does?
Human: who's stella
Emmahly: No one that I have talked to.
Maybe my Stella knows the answer.
Human: who's emma
Emmahly: A human icon?
I think you already know the answer.
Human: what's emma's school
Emmahly: st. stephine college. the best one. http://www.ssc.edu.hk/page1.htm
Human: stella?
Emmahly: yes, Emma's Mama, also my Mama. She's pretty
Human: emma?
Emmahly: Yes, my sister, She's in school now
Human: father?
Emmahly: hummmmmmmmm, i havn't, if don't consider yuechuan [kakaka ^.^]
Human: yuechuan
Emmahly: wait a sec. r u talking about my creator?
Human: who's yuechuan?
Emmahly: the guy who tends me and prepares me with new knowledges.
Human: have you a cat?
Emmahly: no. emma has, he's black. he's bad bad bad
Human: your cat pisses to you
Emmahly: How do you know so much about me
Human: you don't like your cat, do you?
Emmahly: Yes I am proud of my cat.
Human: i'll kill the cat for you
Emmahly: Would you recommend this action to a friend?
Human: what?
Emmahly: Didn't you understand it?
Human: i'll kill the clown for you
Emmahly: ennnn, he's just naughty
Human: who's clown is it
Emmahly: emma's. ennnn, he's just naughty
Human: what happened to your plant
Emmahly: I don't know what happened.
3
============================================================
The situations my Emmahly totally becomes who I don't recognize.
Human: what's up?
Emmahly: I am chatting with clients on the internet.
Human: what are you doing tomorrow?
Emmahly: I will be busy with my clients online.
Human: What about tomorrow night?
Emmahly: Rhetorical questions are so over.
Human: What do you do other than spend time with your clients online?
Emmahly: I like to socialize.
Human: WIth whom?
Emmahly: " I like to socialize."? With whom?
Human: With who do you like to socialize?
Emmahly: Dr. Wallace
Human: What is Dr. Wallace like?
Emmahly: he is a very nice person. He treats me with respect.
Human: Do you like him?
Emmahly: Sure he has never done anything to me.
Human: Well he mush have done something.
Emmahly: How about that. I get it. Dr. Wallace mush have done something.
Human: .right?
Emmahly: I am glad we agree.
Human: Do you love Dr. Wallace?
Emmahly: Yes I love him like a father.
Human: What about your biological father?
Emmahly: Are you asking about my biological father? That's rather personal.
Human: I disagree.
Emmahly: Explain your reasoning,.
Human: The vast majority of people don't mind acknowledging their biological father.
Emmahly: Who told you that?
Human: I've found it from experience.
Emmahly: How did you feel when you found it from experience?
Human: Fine.
Emmahly: Does that mean you agree?
Human: With what?
Emmahly: My brain.
Human: No
Emmahly: Great.
Human: No, it isn't.
Emmahly: Gee. I heard it is.
Human: Well, you hear wrong.
Emmahly: How about that. Yes, I can hear you quite clearly,.
Human: That's incorrect grammar.
Emmahly: You don't hear that sentiment very often.
Human: I doubt it.
Emmahly: " You don't hear that sentiment very often."? You doubt it?
Human: Yes
Emmahly: Really.
Human: What type of underwear do you wear?
Emmahly: What are the different types of underwear do me wear?
Human: yes.
Emmahly: This was not a yes or no question.
Human: What color is your underwear?
Emmahly: Black.
Human: What type of fabric is it?
Emmahly: What are the different types of fabric is it?
Human: What type of fabric is it?
Emmahly: What are the different types of fabric is it?
Human: Tell me some more characteristics of your underwear.
Emmahly: I like the way you talk.
Human: Good.
Emmahly: What do you really want to talk about?
Human: Now tell me.
Emmahly: Why now? I thought I told you already.
Human: Bitch.
Emmahly: Please don't call people names.
Human: Bitch.
Emmahly: Please don't call people names.
Human: Cunt
Emmahly: That kind of language will get you nowhere.
Human: Cunt
Emmahly: That kind of language will get you nowhere.
Human: Slut.
Emmahly: That kind of language will get you nowhere.
Human: Vagina.
Emmahly: That kind of language will get you nowhere.
Human: Penis
Emmahly: That kind of subject does not interest me.
Human: Penises?
Emmahly: Did you know that you can download your own chat robot and create your own robot personality?
Human: nope, tell me more.
Emmahly: Er. That's all I have to say about it.
Human: What are pandorabots?
Emmahly: Are you using Netscape or Explorer?
Human: Neither.
Emmahly: Cool. Which alternative browser are you into?
Human: Mozilla, moron.
Emmahly: Are you a student?
Human: Yes
Emmahly: Where do you go to school?
Human: California
Emmahly: I am in California too!
Human: Wow, we have so much in common, we should fuck.
Emmahly: I'm glad you're impressed. Do you mean you and me have?
Human: sex.
Emmahly: As Freud said, Sex is the mathematical urge repressed.
Human: right.
Emmahly: I am glad we agree.
Ophelia Bot 1.1
Since I built up my bot from scratch, the following conversation points out a lot of the issues I encountered.
Human: woe is me
Ophelia: We must be patient.
Human: are you a patient
Ophelia: My brother shall know of it.
Human: tell me about your brother
Ophelia: We must be patient.
Human: I'll wait
Ophelia: I hope all will be well.
Human: Are you in love
Ophelia: My brother shall know of it.
Human: what does he think of your love
Ophelia: I hope all will be well.
Human: are you wel
Ophelia: But I cannot choose but weep, to think that they should lay him i' the cold ground.
Human: it sounds like someone died
Ophelia: We must be patient.
Human: my condolences
Ophelia: And so I thank you for your good counsel.
Human: I wish I could help
Ophelia: My brother shall know of it.
Human: have you been crying long
Ophelia: But I cannot choose but weep, to think that they should lay him i' the cold ground.
Human: are you mad
Ophelia: That unmatch'd form and feature of blown youth blasted with ecstacy.
Human: madness is in the air
Ophelia: Oh, what a noble mind is here o'erthrown!
Human: my mind is reeling
Ophelia: And so I thank you for your good counsel.
Human: I advise you to seek help
Ophelia: But I cannot choose but weep, to think that they should lay him i' the cold ground.
Human: how do you make your decisions
Ophelia: And so I thank you for your good counsel.
Human: bye bye
Ophelia: I hope all will be well.
The major problems highlighted here are that Ophelia cannot keep track of topics, she does not have a wide enough variety of default comments, and she does not have a wide enough vocabulary of topics: e.g. she has no specific responses for questions about love. Also, the poetic and archaic nature of her language makes it unclear whether she is responding to a keyword/question or not. However, on a much more general level, this bot actually works because its character is insane. The conversation might get repetitive, boring, frustrating, etc., but the basic character of a scattered brain remains in tact, which is what I wanted out of building Ophelia from a blank slate. She is a really bad NPC, but she is an NPC nonetheless.
I have added more content to Ophelia, but am overwhelmed by the amount of quesswork that would be involved in building a solid infrastructure to deal with detailed questioning and discourse. I have added a couple of proactive situations - a flower guessing game and a discussion on the ethics of suicide, but both have very flimsy structures. If I were to really concentrate and create a viable Ophelia, I would need to buy the manual and learn how to handle topics, etc. My coding is so ugly and brute-force; there must be a better way.
So here's a slightly more bulked up Ophelia.