Lodestone Systems has a number of internal research and development efforts in progress. These primarily fall in the fields of Intelligent Agents and distributed applications, and both their application towards discovering knowledge.
The goals of the Intelligent Agent work are twofold. One, better ways to search, profile, and interact with the information tsunami we are experiencing are being investigated. Active searches need to provide better results, particularly as the amount of information rapidly increases creating result sets that are volumous in themselves. Likewise, the automated regular gathering of information on specific topics from selected sources could benefit from a more intelligent and flexible definition of what those topics encompass and when is a particular piece of information pertinent and when is it not.
Two, autonomous and collaborative software concepts are being explored. As our information needs entail more complex attention, it will become desirable to employ software "research assistants" that have the ability to make certain decisions and pursue certain actions on our behalf. The guidelines by which they operate will need to be more flexible than those found in current software tools. Also, Artificial Intelligence research has shown that even simple general knowledge is very difficult to encapsulate. Approaches utilizing a group of very specialized agents that are also capable of interacting with each other may be a more realistic means of realizing intelligent software tools.
While the main projects briefly defined below can be viewed distinctly, they are designed as part of a collaborative suite that can leverage the operations of the others.
Some of these may become products in the future, or they may show up as technology underlying other products.
- NetFerret - Intelligent Searcher
- This agent is designed to ferret out information on the Internet based on a variety of search paradigms and across protocols. Its role is to locate results for a specific, one-time query without knowledge of where those results will be found. The result is not a list of sources related to the search, but actual information. It incorporates the ability to weight the level of trust of certain data corpora and to have that level of trust reflected in cited results.
(Incidently, the name is derived from a similarly purposed type of software described in the book Earth by David Brin, published in 1990, which takes place 50 years in the future. It is interesting to note that this book offers some good forethought of issues we are now facing on the Internet.)
- NetHarvester - Intelligent Profiler
- NetHarvester is designed to multiply information profiling and gathering capabilities by also incorporating source trust levels as well as gathering related information. The role of NetHarvester is to routinely retrieve information related to a defined profile from specified sources. Source trust and correlation have a distinct effect on the value of information. A piece of information common to 4 trusted sources can be just as informative as if it comes a single very trusted source.
- NetBroker - Intelligent Broker
- The NetBroker agent's role is to interact with external entities to effect an action. This action may be the purchase of a plane ticket or piece of information, or the request for catering services or human research assistance. Designed as a shell which can be filled by a variety of goal modules, such as best deal shopper or an online auction bidder, NetBroker can task NetFerret and NetHarvester to obtain information required to make decisions.