Ben Goertzel’s AGI discussion centered around what he called the "cognitive synergy" between the different processes going on in the brain (declarative, procedural, episodic, sensory, attentional, and intentional) and how each of these processes is connected to a specific kind of learning. The human mind is good at integrating these different systems, which have to interact with each other closely to create the cognitive synergy required for a fully functional brain, and therefore a fully functional AGI. He then went on to explain some of the inner workings of the Open Cog open source cognition engine, and how it uses a common knowledge representation to express all these different types of learning.
Ben gave us some basic examples of how his system derives knowledge. For example, if you give it "I saw the man" and "he wore a red shirt," the system could derive that "the man was dressed in red." He explained how the Novamente and Open Cog cognition engines have a language understanding engine that conducts common sense inferencing using probablistic logic networks, and the concept of "symbol grounding," wherein language is understood by association with things the AI agent has experienced in the world.
Next, Ben showed the audience a demonstration video of one of his virtual pet programs, providing thought bubbles to show how the virtual pets learn, and how the process converts text to knowledge and back. In the video, the dog is taught tricks using a combination of imitation and reinforcement learning.
Ben also announced a new collaboration with Xiamen university – through a grant from the Chinese national science foundation, to apply this virtual work to actual robots.
For his last trick, Ben described how he used his Biomind AI program to invent what turned out to be a nearly 100% predictor for Parkinson’s disease, as well as the first evidence of a genetic basis for chronic fatigue syndrome. Ben’s algorithms have also discovered goodies like which tumor suppressor genes impact calorie restriction on aging (conducted in collaboration with Genescient). He’s also collaborating on a life extension project that’s figured out which genes to modify to double or triple the average life of a fly. Since humans have a lot of the same genes that flies do, there are implications for how this discovery could someday extend human life. These various applications were all created by applying Ben’s AI algorithms to existing data in one form or another in order to determine how the genes related to each other.
Are you intrigued by the possibilities of AGI and AI for curing disease? Let us know here on the blog!