The theater model as software architecture
Sunday 25 February 2007 10:36
Bernard Baars' Theater model of the human mind lends itself well as a software architecture for a human-like mind. It is distinct from any of the software architectures that I discerned before. It's main components are:
Contexts do the actual work of the architecture. The solve problems, plan actions, interpret input, etc. Any context can be a complete application in itself. Contexts are not static. They can be created and be destroyed. Contexts receive their input from outside (perceptual contexts) or from WM. Contexts are connected to WM in an active way. They constantly monitor the contents of WM, to see if it contains parcels that may serve as input to their process. If so, they send a request to the Action Scheduler to get access to the WM. If this is granted, the context can act upon the WM parcel or parcels.
The Input Scheduler receives requests from all contexts to place their parcels onto the WM workbench. Based on some criteria of importance it decides which parcel (parcels?) to place there.
The Action Scheduler does the same. Only the requests are actions, not parcels.
The architecture is interesting in areas where:
- Working Memory
- Contexts
- Input scheduler
- Action scheduler
Contexts do the actual work of the architecture. The solve problems, plan actions, interpret input, etc. Any context can be a complete application in itself. Contexts are not static. They can be created and be destroyed. Contexts receive their input from outside (perceptual contexts) or from WM. Contexts are connected to WM in an active way. They constantly monitor the contents of WM, to see if it contains parcels that may serve as input to their process. If so, they send a request to the Action Scheduler to get access to the WM. If this is granted, the context can act upon the WM parcel or parcels.
The Input Scheduler receives requests from all contexts to place their parcels onto the WM workbench. Based on some criteria of importance it decides which parcel (parcels?) to place there.
The Action Scheduler does the same. Only the requests are actions, not parcels.
The architecture is interesting in areas where:
- Many diverse problems are solved simultaneously
- The agent needs to be flexible and respond to interrupting threats/requests immediately
- Input is highly ambiguous and can be interpreted in many ways (the active contexts); interpretation requires a large top-down component, and this component may change all the time
- Many processes are active at the same time, not necessarily solving problems, but also just performing complex tasks; these tasks should be performed in parallel as well
Archive > 2007
December
September
- 9/22/2007 9/22/07, 12:46 PM - Implementation of the parser
- 9/18/2007 9/18/07, 4:47 PM - Saga - Duitse rock uit de jaren 80
- 9/4/2007 9/4/07, 9:18 PM - Rah band - Clouds across the moon - YouTube
August
July
June
May
- 5/27/2007 5/27/07, 10:06 PM - A web agent that forms a user model
- 5/25/2007 5/25/07, 8:18 PM - I am a strange loop
February
- 2/25/2007 2/25/07, 10:36 AM - The theater model as software architecture
- 2/20/2007 2/20/07, 9:39 PM - A Semantic Web Primer
- 2/7/2007 2/7/07, 10:13 PM - My position on consciousness
- 2/5/2007 2/5/07, 9:31 PM - In the theater of consciousness
Reactions on "The theater model as software architecture"
No posts found
Log in to comment on news articles.