Discussion on Insect-inspired Intelligence and BIC
Second Discussion on Insect-inspired Intelligence & BIC
With the emergence of Life Science in recent several decades, the topic of brain inspired computing (BIC) comes as an significant comparison when considering the contribution of one’s work.
Supplementary Reading
Insect-inspired AI for autonomous robots
Introduction
As an introduction, we may first discuss the basic properties and schemes of the different subjects. The research scheme of BIC will be the first thing to consider, which owning a computation based logic to accomplish different tasks. In the view of computer science, task is on top of anything else, which means the consistency of algorithm and biological logic does not matter. As for computing biology, such consistency contrarily becomes the top priority. With the experimental data collected from different task scenarios, a mathematical model will be established through data fitting. The middle place is BIC, aiming to apply biological principles to computing structures, thus called brain inspired computing.
Due to the development of computational power since 2010s, neural networks gradually become a new focus of the above subjects. Training and implementing these bio-inspired model to fulfil specific goals, such as object recognizing, classification, decision making, etc., is the new trend of research that is similar to the role of evolutionary algorithm about two decades ago. Moreover, the implementation of biological principles are quite creative, decorating the research style of BIC with that of the Tycho and Galileo’s time to physics. As is known to all, such time is the mixture of opportunities and failures. The chances usually behave as the application of certain biological principle or mathematical theory. For example, the spiking neural networks (SNNs) are typical instances of the former and continuous attractor neural networks (CANNs) are better classified into the latter. Meanwhile the failures, or challenges, are thrown to other traditional subjects together with their schemes, and robotics is one of them.
The research scheme of robotics in the mechanical engineering view is pretty traditional. However, BIC has thrown the problem of whydunit to traditional robotics, especially in controlling, decision making, and robotic vision. One must ask himself why the traditional strategies are still important even if others can easily accomplish the same goal through neural networks. Some researchers also choose to follow the trend, which means implementing brain-inspired methods in their research work as well. If the complex neural network model cam be applied to all robots, definitely there is no need to put such enormous efforts into traditional methods. The thing is, for micro (the size of which is 10cm or smaller) robots, there is no sufficient space to apply complex controller, with microcontrollers substituted. Under such circumstances, insect-inspired intelligence is proposed.
The idea of insect-inspired intelligence (we may use III to refer to this term in the followings. Note that such simplification is not official) is quite simple. Similar to brian-inspired intelligence, we now specifically consider the way insects behave and apply it to microcontrollers. Different from the neuromorphic computing ways to enable the application and break through the limited computing power such as memristors, neuro-inspired computing chips, insect-inspired intelligence rises from the power of traditional microcontrollers. One may regard this strategy as a struggle of old times’ shadows. For those believed in it, the reason may be presented as “researching insects can offer us more insight of the biological behaving principles under the condition that the secret of brain are still far from explicit.” Correct or not, it is still the Zen of insect-inspired strategies.
The ultimate goal of BIC and III are nevertheless same, i.e. to fully reveal the secrets of brain. The main difference result from the direction, in other words, the former has chosen a top-down way and a bottom-up way for the latter. BIC will start from basic properties of the brain and then construct a model to carry out different tasks, during which data is the driven force. The problem of data-driven deviations are common under such schemes, and thus is called top-down. III will instead base on traditional methods to merge insect-inspired improvements. Since it seeks to apply insect intelligence as inspiration instead of principle and utilizing the traditional theories. By comparing the real behavior of insects and the behavior of robots, researchers can thus grasp more insight into the biological principles. Furthermore, the differences of traditional theory based results and biological observation results will reveal more significant information about the biology world.
The core idea of CPG and BIC are not necessarily different. With comprehensive view into both parts, one can understand more about his work.