Part 8 :   Conclusion , References & Bibliography

8.1.  Conclusion

The concept of time-delayed neural signalling is biologically plausible[7]. By constructing neural circuits based on time-delayed neural signalling, any type of stimulus can be encoded as a scalar to a "standard" circuit, in such a way that it is possible to deduce intensity, detect change and detect patterns in the incoming stimulus. Some innate connections are incorporated for initial survival. Further learning-by-association is enabled by setting criteria for formation, strengthening and weakening of neural connections. The approach described in this paper has the potential to explain other empirical phenomena related to the human brain.

In a neural circuit based system, both data and action-initiation could be encoded in the connectome of the neural network. The neural circuits and other methods described in this paper could be leveraged for use in software applications related to encryption, cybersecurity, and various other use-cases. Software designed based on such neural circuits will be able to determine context, determine the importance of the current ongoing action in the environment and trigger motor actions accordingly. For example, if virtual “pain” is associated with an undesirable action, (eg: detection of unauthorized edit to a database) , a motor action (eg: email alert) can be triggered.

By providing positive or negative reinforcements , more behaviours can be taught to such a system purely by interacting with it. It should be noted that such systems based on Neural circuits are not good at “brute force” remembering (eg: like recalling a long text sequence); however such neural circuits would excel at integrating sensory information, taking prompt action based on context and then adjusting behaviour based on feedback.

Further work is needed on the following topics:
  • Designing comprehensive neural circuits for visual stimuli processing(eg: processing colour information, feature extraction,etc) and speech recognition.
  • Enhancing/refining the criteria for managing connectome (connection creation/strengthening/weakening)
  • Illustrating how abstract concepts are coded in neural connections and associated with relevant sensory stimuli,and how visual feedback pathways enable invocation of "mind's eye"
  • Incorporating dynamically activated neural pathways in an Artificial Life Form by setting neuron thresholds as a function of a variable (eg:virtual hormone level), which activates dormant neural pathways and simulates biological phenomena such as circadian rhythm, phobias to certain stimuli,etc

8.2.   References


  1.   Dispense with Backpropagation : Artificial intelligence pioneer says we need to start over
  2.   Reflexes in newborns: Neonatal reflexes
  3.   The law of the synapses: punishing the weak to maintain strong synapses strong
  4.   How the body shapes the way we think: Rolf Pfeifer at TEDxZurich
  5.   How will an AI be made to feel pain?
  6.   Motivation and Reward in Learning : Yale University
  7.   Neurons can learn temporal patterns

8.3.   Bibliography


  1.   Robert Sapolsky : Behave: The Biology of Humans at Our Best and Worst
  2.   Aubrey Manning, Marian Stamp Dawkins : An Introduction to Animal Behaviour