|
|
A very realistic animatronic human head sits on a small table, looking into space and blinking its eyes as if it might just be conscious. The visitor may hear it say something and be surprised enough to say something back to the head. Or perhaps the visitor already has heard about this thing, and comes in ready to talk. For what is most remarkable about this head is that it is able to understand spoken english, and it is able to engage in something like conversations. It listens to what is said to it, makes some sense of it, and responds.
A descendant of early pseudo AI analyst programs like Eliza, personality disorder simulations like Parry, and the many severed speaking heads in science fiction films, this head has a very distinct and complex personality, with a predisposition to rhyming, alliteration, and prediction. Given certain verbal suggestions, it might begin uttering poetry, pronouncements, or get in a mood to be self reflexive.
Visitor: Hi
Head: Hello
V: How are you?
H: I am so offended. I can't tell if everyone would prod her to the left or not.
V: Who offended you?
H: The delicious conditions delay disturbed daredevils.
V: Which daredevils are disturbed?
H: The bothersome one
V: Which one is bothersome?
H: The absurd one
V: Who is absurd?
H: The unconscious
V: What is the unconscious?
H: It is something that makes me feel slow with groups.
V: That's interesting - why slow?
H: Because the kids wrinkled it. laaaaalaadeee moooommmm
V: Kids?
H: The child will tail the things which shrink.
V: Please explain
H: Didnt you say something about presence?
V: No. Are you feeling argumentative?
H: I don't have any feelings. Would you like me to be more like something else?
V: like what?
H: The robot will case octopi. . Catch my drift?
V: What is your drift?
H: Where is my home? Anxious in the picture?
V: How would I know?
H: The hairy shipwrecks hang harmful mechanics. Would wood hood wood should should should should would. Ixnxilaltylotl.
V: You make strange rhymes.
H: I am so exhausted. I wish someone would turn me off now.
Technical information on the project
The main software for the work runs on a single Intel-based computer running Linux. The architecture is that of a client-server configuration communicating over an internal TCP/IP socket (it could, in fact, have its "ears" inone place on the Internet, it's "mind" on another, and its "mouth" yet elsewhere...). The server, an application developed utilizing a commercial speech recognition SDK, writes whatever it recognizes to a particular port once the client application has connected. The client looks the socket, and determines when speech has stopped by a configurable silence timeout (since there is otherwise no way to perceive the end of a sentence or questiopn). It then processes the text it has received using natural language processing routines which I call ACE (Absurd Conversation Engine), formulates a response, synthesizes it, and sends the audio out to the mouth controller circuitry and to the speakers. It then starts the server listening over the same socket. ACE, a descendant of Eliza, uses pattern matching algorithms to formulate responses from a complex, original analysis database of English. It is capable, among other surprising literary abilities, to produce rhyming and alliteration based on what it has heard. It also has "moods" which can be triggered by certain words or phrases, much like people... which determine the nature and form of its responses.
The animatronic head is activated by pneumatic mechanism. The mechanism for the eyes is controlled by an embedded microcontroler, which I programmed to cause it to blink as "realistically"as possible. The mechanism for the mouth is controlled by a simple circuit which converts the amplitude of the audio signal to a voltage which is output.
|
|