|
Anthrobotics™
Who Are We? - The Real Short Version
In the '60s, with computers going big-time, many people were sure it was just a matter of time before Strong AI was here. It never happened, and many say it never will. What is called "Strong AI" (fast definition) is the real thing, as in "HAL-9000" or "C3PO" or "Commander Data".
Strong AI = "real" robots; what you think of when you combine the word robot with your imagination. It's robots that think and learn and feel (emotion). And do natural language! Cognition, consciousness, comprehension. Making autonomous decisions in situations both familiar and unique to them. Goals, strategies, plans. Interrelationships with other robots and with people including not only conversation, but also empathy and bonding (eat your heart out, Commander Data).
Anthrobotics has developed a new and proprietary approach toward the Strong AI problem. It's an overall approach that can make use of the hard-won prior AI advances. It isn't mushy psychobabble or vague AI-talk either. We believe we have it down measurably because we have the foundation down in concrete, measurable terms, namely a computer program. We believe the odds are very, very high that our success so far shows we can do the rest.
What is programmed and working successfully so far? Our virtual robot lives in a virtual world of infinite and random variety (no "canned" worlds here which any non-thinking, so-called robot can live in). There are predators and prey. Rocks and trees the robot doesn't want to run into. At any instant in time our robot knows what is important and relevant in its situation. This enables our virtual robot to seek opportunities relative to its goals and to avoid and overcome threats and obstacles by using strategies and plans. It experiences hurt from injury, hunger and tiredness. And if the robot lives to tell about it, it can and will, in natural language, with emotion and facial expression!
Who are we? We're the ones who know how to do all of this and have done most of it.
One implication? An operating system or a compiler with a natural language interface that is empathetic with its user! It's not a question of knowing how, it's a question of resources to do it. Therefore...
What we are interested in is a strategic partnership to expand opportunities for taking advantage of our technology.
Not As Short
Anthrobotics has developed a new and proprietary approach toward the "strong AI" problem (typified by machines like the fictional "HAL-9000" or "C3PO" or "Commander Data"). The software platform ideas used for feasibility studies of this new approach have provided immediate commercial "by-product" potential for intelligent entities in several product domains, including games and other entertainment products as well as in virtual realities.
The developments include a new approach to natural languaging (including metaphorical) by machines. In current efforts at intelligent machine-understanding of natural language, the processing systems have applied things like rules and/or "matching" techniques and/or neural nets. In none of such efforts does the incoming language produce "meaning" within the processing system.
Humans, on the other hand, handle language by producing "meaning". They have mammalian brain systems by means of which they make sense of the world and make autonomous decisions WITHOUT language, including many parallel and interactive sub-systems for things like handling sensory inputs and motor outputs, for recognizing objects, for making internal representations of apparent situations, for selecting goals and plans, for trying out potential actions in their heads, for making decisions, and for learning, even for recognizing and making use of emotional feelings like fear and metabolic feelings like exhaustion. For humans, the meanings of natural language come from the linking of such language to those underlying NON-language autonomous systems so that the language engenders pictures -- feelings --plans --possibilities.
As for humans, so for machines also, at least at a systems level. If there are no underlying mammalian-like autonomous systems, then there can be no human-like language understanding, no human-like language translation. Anthrobotics discovered how to build, digitally and discretely, such underlying non-language autonomous systems in a machine in such a way that natural language could be linked to such systems. When language is linked with these underlying non-language systems of the machine, some hard technical things get easier to do. For example, a machine that knows what circumstances are relevant to a situation or communication and what the
implications are can slice through some of the current complexities of language comprehension.
Anthrobotics machines include a radically-new "concrete" representational field capable of instantly simulating the infinite variety of the natural world by various combinations of a surprisingly small set of "primitive" elements. To determine current and predicted threats and opportunities to the machine, this representational field is constantly compared (in a "parallel" mode) to a relevancy network including all the kinds of "abstract" situations which are "meaningful" to the machine. The machine may also use this feature in a "what-iffing" mode to "preview" the implications
of decisions it is considering.
Eventually, among other things, the Anthrobotics approach will help provide natural-language access to a "friendly" and thoughtful personal communicator -- a personal assistant. That personal assistant would be the customer's interface -- maintaining contact the way the specific customer wants -- with the immense and fast-changing information and technology of the 21st-century. As such personal assistants become more and more affordable, who will not want one?
| 
