Product Overview

We current support two DAI products:

• Alchemy - a parallel/distributed environment
• Goal Mind - a cognitive modeling environment

Using Alchemy/Goal Mind, a cognitive model can be designed and built using any number of copies of a number of different embedded inference engines (such as production systems, semantic networks, etc.), and then, distributed over any number of processing elements in a cluster or grid. Once the theoretical aspects of the model have been verified, Alchemy/Goal Mind provides all the necessary tools to integrated model components into a multi-agent application that can itself be easily integrated with existing software and/or hardware systems.

Alchemy

Alchemy is a generalized parallel/distributed environment that solves a number of concurrency, setup and security problems that exist in environments like RPC or CORBA. It is based on a highly-efficient low-overhead client-server communication model that abstracts socket connection complexities by using a built-in connection name space. It improves parallel throughput by supporting asynchronous messaging between processes while still maintaining the security and reliability of connection-oriented traffic. It contains it own multi-level security protocols for message traffic that can be easily controlled within an Alchemy application to provide 'just-enough' security, allowing an application design to be tuned to support fast but secure IPC.

Alchemy supports the construction of multi-agent applications from a group of processing nodes and client-server connections. The input, output and process of a node are multithreaded to avoid deadlocks and improve performance on processing node that support SMP and SMT. Using GUI development tools, any number of nodes and node connections can be defined as a graph of processing element distributed across any number of processors. The distribution of an Alchemy application can be dynamically altered at any time from outside or within the application.

Goal Mind

Goal Mind is a generalized cognitive modeling and application development environment using Alchemy. It supports a number of existing inference engine nodes (called RIMs in the Goal Mind theory) that can be combined with other interface and coded-response components (IFNs and CRMs) to build multi-agent based cognitive models. The focus of Goal Mind is on:

1. defining a unified intra-agent and inter-agent communication approach (based on brain areas and social groups)
2. providing a set of reusable RIM, CRM and IFN nodes with integrated knowledge support
3. providing tools to support general knowledge management, model construction and debugging, and realization of models in Alchemy.

Existing IFNs also support integration of a Goal Mind models and applications to other large existing systems; for example, a Goal Mind language model is currently being integrated with the CMU Sphinx and Festival Speech Synthesis systems.

Goal Mind provides built-in support for knowledge management using a distributed relational database (currently PostgreSQL). This allows Goal Mind RIMs and CRMs to access LTM (Long Term Memory) via database lookups, thus, supporting either the sharing and non-sharing of knowledge between RIMs and CRMs. Goal Mind also support access methods to external knowledge bases; for example current components provide access to on-line web resources and WordNet 3.0.