Building synthetic brains capable of human level discovery and invention...

Summary - IEI builds unique neuro-control systems that enable robots to ad lib tactics and strategies that exceed their original programming. Fully capable of autonomously learning from its own mistakes and successes, our revolutionary neural network paradigms allow complex robots to learn completely from scratch. In a matter of minutes, the equivalent of 'cybernetic road kill' can learn to walk, recover from various mishaps, or accomplish some broadly defined mission. While such creative robots are strictly experimental at this stage and under development for the Air Force Research Laboratory, we can tailor similar systems for commercial and private application.

Details - For the most part, modern robots designed for industry and the military are termed reactive.  They simply sense various scenarios within their environment and then recruit the necessary behaviors, in the form of  computer code or a deterministic neural network, to react to such situations. However, the kinds of robots that we frequently see portrayed in science fiction are called deliberative, since they appear to accumulate world models and ponder such models when deciding what to do next. Creativity Machines are the natural way to implement such contemplative control systems aboard robots, since the imagination engine may review a wide variety of action plans, while critic networks may select the strategy most likely to meet the broad objectives of the system. Furthermore, if the Creativity Machine is STANNO based, it can learn to perform various feats from scratch, using totally untrained artificial neural networks. 

In the figure to right, for instance, a complex hexapod robot uses its onboard sonar to judge its forward progress as its Creativity Machine based control system experiments and cumulatively learns how to walk. Later, this very same Creativity Machine can enter into a cycle of experimentation and learning to perfect other types of motion strategies (i.e., backward, right turn, and left shuffling motion). In effect, this neuro-control system emulates the thalamo-cortical loop of the brain, wherein the cortex imagines various courses of action through noise stimulation, while the thalamus narrows down the range of possible scenarios prior to motor cortex implementing the most appropriate of these plans. ...In IEI terms, the noise-activated imagination engine corresponds to cortex, the critic net to the thalamus, and the feedback loop to the reentrant connections connecting the two brain centers.

Taking full advantage of IEI's SuperNets™, robotic brains may autonomously knit themselves together to accomplish even more complex tasks, such as navigating buildings or terrain, or how best to outsmart an opponent system. Such self-organizing brains may effectively go to sleep and rehearse their mission objectives within robotic virtual reality simulations. To see a robot, starting from a state of no learning whatsoever, and learning to walk, see IEI's tabula rasa learning in robots.

For more information on IEI's revolutionary robotic AI schemes, contact Dr. Stephen Thaler.