This is an initial description of an extension to the Brain Simulator III. Ongoing development and experimentation will contribute to more details.
Overview
The Brain Simulator relies on its ability to handle large amounts of data to draw conclusions and implement Common Sense. In order to jumpstart the knowledge base, an agent within the Brain Simulator will:
- automatically recognize when additional knowledge is needed
- then go to one or more online information sources to retrieve it
- add the retrieved information to the UKS
Experimentation for this process has begun in the module “OnlineInfo” which can reach out to any of eight data sources to retrieve information. These include ChatGPT and an online kids dictionary which contains simple definitions and simple images for many objects. Item 2 (above) works reasonably well either through APIs provided by the data sources or though webpage scraping. Items 1 and 3 are hard-coded proof-of-concept stubs.
Issues to be Addressed
How can the agent select which information to search for?
How is retrieved data parsed and formatted so it can be put into the UKS?
How will the agent handle conflicting and/or erroneous data?
