Artificial Stupidity/ Artificial Intelligence

Artificial Stupidity/
Artificial Intelligence

August 1-4, 2002

This event is a part of The Banff New Media Institute’s
Human Centered Interface Project.





Conference Research Report Prepared By:
Nicholas Zaparyniuk

---

Artificial Stupidity/Artificial Intelligence

Thursday, August 1st.......................................................................................3
Welcome and introductions............................................................................3
Imagining Machine Intelligence......................................................................4
Phoebe Sengers, Herve Fischer, Rita Lauria, Ben Whittam-Smith ............4
Getting Down to Data: Large Scale Data Sets, Intelligent and Subtle
Narratives ......................................................................................................9
Andrew Salway, Zack Simpson..................................................................9
Imagining and Creating Intelligent Characters .............................................11
Ken Perlin, Adam Frank, Mirjam Eladhri, Ralph Guggenheim..................11
Friday, August 2nd .........................................................................................17
Morning Summary .......................................................................................17
Sara Diamond ..........................................................................................17
Learning through Machines, Machines that Learn .......................................18
Michael Gordon, Rowena Goldman, Michael Buro ..................................18
Organic Lives, Other Than Our Own ...........................................................20
Nell Tenhaaf, Christan Jacob, Demetri Terzopoulos ................................20
Opening of "Mapping our Territories" in the Walter Phillips Gallery .............23
Are You Human? Modeling Humans--Thinking with and For You: Agent
Technologies and Neural Networks .............................................................24
Catherine LaBore, Katherine Isbister, Marina Zurkow and Scott Paterson,
Tom Donaldson ........................................................................................24
Writer’s Talk and Artist Celebration as part of the Summer Arts Program! ..28
Steve Grand, David Rokeby.....................................................................28
Saturday, August 3rd .....................................................................................33
Recap of the day..........................................................................................33
Intelligent Games, People, Machines, —Playing By Whose Rules?............34
Jonathan Schaeffer, Celia Pearce, Eddo Stern, John W. Buchanan........34
As We Go Marching, Marching, Robots And Performance ..........................39
Adrianne Wortzel, Christopher P. Csikszentmihalyi .................................39
Conversation Systems, Smart and Stupid Language Analysis and Data
Mining ..........................................................................................................41
Randy Goebel, Sheelagh Carpendale, Sara Diamond Richard Lachman
and CodeZebra Team ..............................................................................41
Games Night! Public Discussion Of Popular Games And How Characters
And Intelligence Work In These. Followed by Games, by XBOX................44
Ralph Guggenheim, Steve Boxer, Zack Simpson, Mirjam Eladhari, XBox
Representatives .......................................................................................44
Sunday, August 4th: Research Day .............................................................46
Review Of Key Ideas, Key Research Questions, Project Ideas ...................46
Artificial Stupidity/Artificial Intelligence August 1-4, 2002
- 2 -

---

Artificial Stupidity/Artificial Intelligence
The Banff New Media Institute at the Banff Centre for the Arts
August 1-4th 2002

From August 1st to the 4th 2002, the Banff New Media Institute (BNMI) at the
Banff Centre for the Arts hosted and facilitated “Artificial Stupidity/Artificial
Intelligence” (AS/AI) a symposium on the creative possibilities and limits (if any)
of Artificial Intelligence (AI). Lead by Sara Diamond, the executive director of
Media and Visual Arts, “Artificial Stupidity/Artificial Intelligence” addressed the
questions of the capabilities of affective computing in our ability to shape
qualitative experiences that represent the fuzzy line between computation and
intelligence. Begging the questions of how much a machine can or cannot do
and how much is left to our own representation devices, AS/AI, looked at the
boundaries of computation and our own interpretation of exactly what intelligence
is. With current applications in the analysis of large bodies of data, game play,
understanding speech, and the mapping of learning, how far we can take
computation in our plight to understand our own cognitive and social systems is
unknown. Amongst others questions, this symposium addressed the larger
question of “are artificial intelligent systems only as stupid or intelligent as we
choose to make them”. Bringing together artists, social scientists, researchers,
game designers, and A.I/ A.L. experts, the forums of AS/AI focused on the wide
range of applications, roles, and possibilities of artificial intelligence.

Thursday, August 1st

9:00am – 10:00am
Welcome and introductions
Sara Diamond – Executive Producer, TV & New Media, MVA, The Banff Centre

Sara Diamond introduced the event with a brief history of the Banff New Media
Institute and some of the objectives they hoped achieve during the summit. With
a history since 1994 of bring together people from a wide range of disciplines to
look at cross-disciplinary issues and questions in new media, the Banff New
Media Institute is special place for collaboration and communication. Bring
cultural science, humanities and computing science together to promote dialog of
the issues of practice, philosophy, and hard science on how we can collaborate
together and work through the issues of new media. With new media being a
large complex area, Diamond defined it as an umbrella term to address a diverse
area including nanotechnology, genetic engineering, computing science, gaming,
and interactive experiences. With all those disciplines addressing the cultural
marriage of humanities, social sciences and computing. With a brief background
on the centres previous symposiums including “Unforgiving Memory” and “Big
Game Hunters”, Diamond outlined how the dialog of these symposiums brought
about questions of practice and philosophy. From these forums the questions of
human machine interactions, avatars, virtual worlds and machine intelligence
presented itself.

Artificial Stupidity/Artificial Intelligence August 1-4, 2002
- 3 -

---

With the overall theme of the conference of Artificial Stupidity or Stupid A.I., the
aim is to understand more clearly what machines can do and how it can be
quantified. Looking at the questions of the human experience and machine
intelligence, the conference looked at the range of topics that populate A.I.
Emotional and affective computing, machine intelligence in games, speech
analysis, data, and our understanding of machine consciousness, the symposium
aimed to bring about a dialog of the blurring between human experience, Stupid
A.I. and A.I.

10:00am – 12:00am
Imagining Machine Intelligence
Phoebe Sengers, Herve Fischer, Rita Lauria, Ben Whittam-Smith

Phoebe Sengers, Assistant Professor, Cornell University

Phoebe Sengers, an assistant professor at Cornell University, and the first
speaker of the forum generated a buzz with her talk “Interpreting AI: Some
Heuristics”. With a background in artificial intelligence, human computer
interaction and cultural theory, Sengers talk focused on her interest in what A.I is
and how it could be better. With a broad understanding of the cultural roots of
A.I, Sengers views A.I with many questions as to the function, impact and future
roles.

In observing the behaviours and the multifaceted nature of both animal and
human interactions with the outside world, Sengers noticed that our ability to
simulate these entitles was somewhat limited. With the simplification and
breakdown of entities as parts, rather than addressing the interaction/entity as a
whole, we were missing the richness of the original being. Sengers went on to
say that the mere modeling of an interaction or agent in the normal way of
programming A.I always ended with a simplistic model of the interaction which
subverted what we were originally interested in the first place. This way of
modeling behaviours or interactions, which is prominent today, is not sufficient to
capture the essence of the being. These flaws lead Sengers to start critically
evaluating A.I in its application and synthesis in the realm of interaction.

What Sengers found in her work was that the current approach to programming
A.I was flawed. That breaking behaviour or interactions of an entity into
components does not represent the organism or its interactions. With the
scientific process of dividing everything into specialized quadrants and trying to
understand each really well, we miss the whole picture. With the adage that
“Sometime the parts do not make the whole”, Sengers went on to say that when
we focus on the parts we miss what made the whole so interesting and
complicated. So with setting the stage for the flaws of current A.I interaction
programming, Sengers posed the question, “is there a way to do A.I in a way that
respects and represents the whole?”

Artificial Stupidity/Artificial Intelligence August 1-4, 2002
- 4 -

---

Sengers answer to this question was that we should not focus on the models that
try to mimic human behaviour or action, but to bridge the formal models of the
code and the human being who is interacting with the system. Outlining the
heuristics of how to build the bridge between humans and computer models
Sengers gave three main points.

The first point was to instead of trying to represent and model the complexity of
an entity, we should be trying to trigger complexity from the user. This way we
would create an experience of complexity that is user centered. An example of
this would be to communicate to the user the idea of what cat ness is all about,
instead of trying to model the behaviours of a cat. This way of programming
complexity allowing the user to interpret the information and build the complexity
of cat through there own devices. In reference to Agent architecture, building
characters that interact through narrative with the user, the activities of agents
and their complexity is in the mind of the user, which contributes to a fuller
experience.

The second point Sengers made was instead of trying to represent this
complexity we should be trying to bootstrap from it. With live triggered systems;
we can represent complexity from the users interactions with the computer.
Because the computer reacts to the users input in real-time, it seems to
represent a complex system. This feedback system is a way to get complexity
into the system without programming it at all.

The last point made was that we should look at the means of the user, as oppose
to the information that is contained in the system. This idea based on affective
computing, aims to extract the feeling of the user to drive the system. Focusing
on affective information as apposed to affective feelings.

With these outlined, Sengers finished her talk with the suggestion that A.I. should
be H.C.I. With these two linked we could build continuously evolving systems
built on the human aspects of the human computer interface and it’s interactions.

Hervé Fischer, Chair Daniel Langlois for Digital Image and Sound, Concordia
University

Herve Fischer, the second presenter of the day, in his talk “Human Memory,
Computer Memory” looked at machine and human memory, the complexities,
fallacies, and, future connections.

Starting his talk with the notion that we must look to the past to rediscover what
memory is, Fischer outlined how memory in the human sense, is completely
different from computer memory and that these systems are no were near
equitable.

Relating that something as simple as the power of smell to trigger memory, we
see that memory can be triggered at any instance in or out of the context of the
Artificial Stupidity/Artificial Intelligence August 1-4, 2002
- 5 -

---

users experience or environment. This power within human memory can bring
forward long past experiences and memories that were not consciously brought
forward, but triggered. This complex system is something that we don’t
understand and is not possible in a computer due to the subjective nature of how
the memory is brought about. This idea of unconscious memory and
unconscious retrieval is something that is problematic in computational/machine
memory models.

Commenting on Freud’s idea that memory is something that never disappears,
Fisher went on to say that our unconscious memory stores everything that we
experience in a system that is not understood. This idea of unconscious memory
is very important in the relation to computer memory, in that we don’t understand
how it presents itself or how it could be repressed or related to computational
models of memory. With most human memory having irrational links that can
present itself at anytime anywhere, the computational model of memory has no
where near the ability or complexity to mimic or compete with this system.
With machine memory being a very simple system, superficial in comparison to
human memory, we see that the power of machine memory is quite crude in
comparison.

Moving to the relationships of the vulnerability of machine memory to human
memory, Fisher talked about the dangers of the transfer of information from
traditional systems to machine memory.

With the Freudian idea that human memory cannot be erased, we see that in
relation to machine memory, which can easily be lost or erased, there is
vulnerability. Relating this idea in his book “Digital Shock” Fisher relates that the
“law of memory” is that the more power and sophisticated the system is, the more
vulnerable the memory is. Using the analogy of a painted picture on a cave wall
that could last 30000 years or more, Fisher outlines how digital memory, in the
sense of CD ROMS and hard disks, storing the same information, can be easily
lost or rendered unusable. Even with extremely sophisticated systems there is
no means to record everything.

With millions of dollars and resources going into digitizing information, we should
be able to record memory. But the fallacy of this notion is that we cannot record
everything. An example of this is the evolving architecture like the web. As we
can see in the short life of the web, we cannot capture the process. As
WebPages are built and improved on, the old versions disappear. In this
example we see the history of websites disappearing everyday with its evolution,
with no memory of how it evolved. Even with the improvement of hardware, we
see that the devices that are used to store memory are obsolete, far from the
paradox of the paintings on the wall. This paradox of vulnerability with
complexity, warns Fisher, has to be brought to the forefront of our minds when
dealing with memory.


Artificial Stupidity/Artificial Intelligence August 1-4, 2002
- 6 -

---

Rita Lauria, Research Associate, Media Interface and Network Design, Labs of
Michigan State University

The third presenter of this group was Rita Lauria, a research associate of the
M.I.N.D labs at Michigan State University. In her presentation “M.I.N.D. Lab” she
talked about the work at M.I.N.D in the areas of human-computer interactions
and intelligence augmentation. With a focus on two main areas of human-
computer interaction research and 3-D virtual environments, Lauria, specifically
looks at the design and philosophy of virtuality.

Asking the question of “what does virtuality imply”, Lauria, outlined that virtuality
implies, a simulated reality that has a distinct line between what is real and what
isn’t real. This distinct line or the “virtual barrier”, as Lauris calls it, is the place
where simulation and reality begin to fuse. With this ability to blur the real and
virtual, we see that virtuality has the ability, not only generate known realities, but
new realities.

The system of virtuality, however, is often misrepresented and misunderstood as
virtual reality. This common misconception however can be easily put to rest
with the distinction that virtually does not represent reality, as VR does, but
interprets and restates reality.

Exploring this idea further, Lauria moved on to the current media environment of
high bandwidth telecommunications. Seeing the ability of our interfaces to
interpret and restate reality, Lauria expressed that there is still dependence on
how advanced the system is. Looking at four dimensions of telecommunication
interface evolution, Lauria outlined each.

The first dimension is the level of mediated environment, which refers to the
degree in which the users body is hooked up to the telecommunications system.
The second is the level of interface intelligence, which is the interfaces ability to
interpret, adapt, and respond to the users behaviour. The third is the ambiguity
of access, which is making make our interfaces more mobile. And the last
dimension refers to sociability, or the number of users that the interface supports
and the quality of interactivity contained within.

These interfaces are a part of what Lauria calls participatory systems. These
systems have the ability to generate and interpret a representation that is totally
different from reality. Virtuality in these systems is used to make the
representations seem real. With that distinction, Lauria went on to say that
virtuality is the conceptualization of “seeming”. That the realities and
representations are projected and perceived, or “seem” real. This allows the
system to be prescriptive with their representation, rather than descriptive.
Looking at the inclusive nature of virtuosity systems, we see that inclusion
becomes a part of the system. We are then in the information and a part of the
information. This then influences a paradigm and possible cultural shift. With this
Artificial Stupidity/Artificial Intelligence August 1-4, 2002
- 7 -

---

in mind and asking the question of where we are in the idea of virtuality, Lauria
presents six conceptualizations that are present in literature that support the idea.
Namely social richness, realism, transportation, immersion, the social actor within
medium, and the medium as social actor.

Finishing with the important idea of presence in virtuality, Lauria defines it as the
ability to project ones mind through the use of media. This “allusion of
perception” is the extent in which the person fails to perceive the medium during
a technologically mediated experience. This idea, Lauria presents, is observable
in how we respond to a comedian on television. With little regard for the medium,
we respond to the comedian as a presence.

Concluding that the condition of virutality is hear to stay, Lauria, asks the larger
questions of; is there any universal truths that can be addressed and interpreted
through this medium.

Ben Whittam-Smith, Notting Hill Publishing Ltd,

The last presenter of the morning, Ben Whittam-Smith in his talk “Concerning
Intelligence: Consumer Perspectives,” talked about leaning from consumer
reactions to A.I. systems and their applications in entertainment. Looking at
three products that Whittam-Smith has worked on, he presented what was
learned through the implementation of these systems in the area of human
computer interactions and empowering interfaces.

Starting with “Dance DNA”, an application that encourages the user to breed and
make 3d objects and animations that dance to music in real time, Whittam-Smith
outlined how he used genetic algorithms and neural networks to create a type of
interface that allowed the user to make these models with little skill. With a very
simple interface, users could create, mutate, and populate the space with new
objects and animate them to music in little time.

The consumer response to this application was overwhelmingly positive. With
consumer points that the application was compulsive and addictive, and that they
could see their music anew, the project was an overwhelming success. With
other aspects of the ability for performance, there was a large response for the
further development of this project. Other aspects of the consumer response
however, lead to surprising revelations.

Finding that users were quite happy to attribute intelligence to the system and
that it didn’t really make a difference if the system learned from the user or not,
lead Whittam-Smith to certain conclusions. Also that the perceived intelligence of
the system was based on a case-by-case basis, lead Whittam-Smith to think that
intelligence may be a localized occurrence. Using this information to build his
second project, Whittam-Smith again looked at a system that reacted to music.

Artificial Stupidity/Artificial Intelligence August 1-4, 2002
- 8 -

---

With the role/goal to use A.I. to create empowering interfaces, in the second
project “the blue player: hitting the spot” a bluebug in the interface reacted to
music. Displaying different graphic images based on the sound input, a series of
sequences based on the music would be displayed. This system again facilitated
the consumer to attribute the application with intelligence. With little regard in
how it worked, the interface had the attributes were the consumer felt included in
the interface.

Looking that the third product “emergence”, built using chatbots, Whittam-Smith
looked at stimulus response case based reasoning. Wanting to see if what was
learned from the previous projects was still relevant, “emergence” tried to trigger
complexity through the user. Using mobile phones because of their ubiquitous
nature, their ability to act as both transmitters and receivers and because they
are always on, made for an easy medium to do the test. What was found was
that trying to use generated content from a chat bot to stimulate or trigger content
from the users stimulated a very high consumer response and engagement.

Finishing with the point that social agents can stimulate and coordinate
information flows, Whittam-Smith warned that they must know their place. With
the ability for the computer and interface to empower the user, the power game is
something the computer has to loose.

1:00pm – 2:00pm
Getting Down to Data: Large Scale Data Sets, Intelligent and Subtle
Narratives
Andrew Salway, Zack Simpson

Andrew Salway, University of Surrey

Andrew Salway, a researcher in intelligent multimedia information systems in his
talk, “Intelligent multimedia data systems: adding structure and meaning to data”
talked about information extraction techniques for multimedia.
Exploiting collateral text that is associated with many complex multimedia
artifacts, including paintings, dance, scientific literature and film, information can
be presented in descriptive form that allows the information to be archived,
accessed and manipulated.

Pointing out that the progress in the direction of intelligent data systems is based
in the idea that the machine must have a means to understand the structure and
meaning of artifacts, Salway goes on to say that this starts with giving the
machine the means to add meaning behind the artifact. Illustrating this with an
example of how a movie is organized into a sequence of moving images and
frames with a specific name, the machine sees this as a file with a name and
size. Both of these views also contrast the human’s view of the movie as
characters, theme and story. These attributes however, should be included in
the machines structural meaning of the file so that interactions with the media
follow on many levels. Moving on to another example, Salway sees the file
Artificial Stupidity/Artificial Intelligence August 1-4, 2002
- 9 -

---

structure on DVD’s, where scenes can currently be accessed, in the future being
able to access more complex information like scene action or theme.

The role of A.I in this process is to move the machine into appraising what the
human experience is, in these contexts, and move some this complexity into the
machine.

So with that vision, Salway asks the questions; “where are we going to get the
structure and meaning from.” Answering this question, Salway approaches this
from a linguist’s point of view. Exploiting collateral text, and looking at what
people are saying and writing about the multimedia as an asset, these systems
could be empowered to do a deeper analysis of the asset. There is however still
a divide between what the text says and the meaning or experience of the
multimedia.

Moving onto working with movies with audio descriptions, Salway asks the
question on whether text descriptions can be linked with narrative so that the
user gets the full experience of the action and narrative of the story along with the
dialog? And moving to the next level of the question, Salway wonders if there is
a way of linking the text with the movie by some form of machine linkage that
understands narrative?

With no current answer for these questions, Salway gave an example of audio
description by showing a snippet of the film “ The English Patient”. With the
description on, it was seen that the potential of audio descriptions to follow and fill
in narrative, apart from the visual, was sparse but effective. The problem now
comes with having a machine understand these connects.

Currently looking at TIWO (television in words). Salway, has analyzed temporal
information and “emotional words” in 85,000 word copies of scripts (12 movies).
Looking at emotion words as a part of a review of information extraction
technology, Salway sees how these words map to the action and description.

Finishing with a “wish list” of future data systems, Salway looks to a future where
complex video queries can be made on any data field of theme, scene,
information, or even character clothing. Also having the ability to summarize the
action, theme, or through points of a film, with a “why” button to bring the user up
to speed, is a future that Salway hopes to build and contribute to.

Zack Simpson, Artist and Researcher

Zack Simpson, the former Director of Origin entertainment and currently an
interactive instillation artist, talked about game design, the taxonomy of game A.I,
and the role of A.I in character and narrative.

Starting his talk with saying that most game developers know little about A.I,
Simpson, related that the priorities of most game designers are graphics. Giving
Artificial Stupidity/Artificial Intelligence August 1-4, 2002
- 10 -

---