|Applications of a Behavioral-Developmental Stage Model to Intelligence in Animals, Humans, and Androids|
|Saturday, May 26, 2018|
|5:00 PM–5:50 PM |
|Manchester Grand Hyatt, Coronado Ballroom DE|
|Area: DEV/EAB; Domain: Translational|
|Chair: Michael Lamport Commons (Harvard Medical School)|
|CE Instructor: Michael Lamport Commons, Ph.D.|
Androids may be based on the behavioral developmental and evolutionary stage model: An Android is "computer based" synthetic organism designed to act like a human. We have created a higher order neural network that thinks, perceives, understands, predicts and manipulates better than insects. Our goal is to emulate the behavior smartest people as measured by their behavioral-developmental stage. Operant conditioning is based conditioning based on four instances of respondent conditioning: 1) respondent conditioning case pairs the reinforcer with the eliciting neural stimulus for the operant response. 2) Pairing of the now salient neural stimulus that elicits the operant with the environment event. 3) Pairing of the environmental event with the reinforcing stimulus.4) pairing of the stimulus elected by the drive with the reinforcement event, changing the strength of the reinforcer. Because of the simplicity of the calculations. These neural networks should be faster and smaller. The first three developmental and evolutionary behavioral stages are Order 1 tasks that are addressed with automatic unconditioned responses; Order 2 tasks include classical conditioning but not operant conditioning. Order 3 tasks coordinate three instances of these earlier tasks to make possible operant conditioning. Neural networks operate at this order as well as some invertebrates and all insects.
|Instruction Level: Intermediate|
|Target Audience: |
People interested in Computer models of behavior, androids and evolution of "smarts" in animals. No experience or knowledge of computer is required. Some background of Respondent and Operant Conditioning is required.
|Learning Objectives: The students will learn about the evolution of animals "smarts" that predict how much reinforcement they obtain by successfully obtain by doing more difficult tasks. They will learn about the evolution of operant conditioning from respondent conditioning. They will learn how to make androids that are equally smarter or smarter than humans.|
|The First Three Developmental and Evolutionary Behavioral Stages|
|MANSI J SHAH (Dare Association)|
|Abstract: The Model of Hierarchical Complexity is a behavioral model of developmental and evolutionary stage, based on task analysis. Tasks are ordered in terms of their hierarchical complexity, which is an ordinal scale that measures difficulty. Successful performance on a task is called the behavioral stage. This model can be applied to non-human animals, humans as well as to androids. Using data from simple animals and more complex one, we will describe the three lowest behavioral stages and illustrate them using the behaviors of a range of simple organisms. For example, Order 1 tasks and performance on them are addressed with automatic unconditioned responses. Behavior at this order includes sensing, tropisms, habituation and other automatic behaviors. Single cell organisms operate at this order. Order 2 tasks include these earlier behaviors, but also include classical conditioning but not operant conditioning. Animals such as some simple invertebrates have shown classical conditioning, but not operant conditioning. Order 3 tasks coordinate three instances of these earlier tasks to make possible operant conditioning. Neural networks operate at this order as well as some invertebrates and all insects.|
|Respondent Conditioning Based on Adaptive Neural Networks|
|SIMRAN TRISAL MALHOTRA (Dare Association), Michael Lamport Commons (Harvard Medical School)|
|Abstract: Adaptive neural networks can be constructed from four cases of respondent conditioning. Respondently based neural networks reduce error and act as amplifiers using the all-or-nothing method. Within each module, a “0” means no Stimulus A and the “1” means an occurrence of Stimulus A. To obtain an output, one multiples either a “0” or a “1”. The first respondent conditioning case pairs the reinforcer with the eliciting neural stimulus for the operant response. The pairing strengthens and makes salient that eliciting neural stimulus. The second case is the pairing of the now salient neural stimulus that elicits the operant with the environment event. The third is the pairing of the environmental event with the reinforcing stimulus. The fourth is the pairing of the stimulus elected by the drive with the reinforcement event, changing the strength of the reinforcer. The network should be able to adapt to its environment of stimuli and better processes the information. It is shown that there are four modules, each representing a responding conditioning case. Because of the simplicity of the calculations, neural networks built in this manner should be faster and smaller.|
|Androids Based on the Behavioral Developmental and Evolutionary Stage Model|
|ANISHA BAIDYA (Dare Association)|
|Abstract: An intelligent agent, or Android is “computer based” synthetic organism designed to act like a human. Researchers are working on developing computational models of human behavior. They are very far from accurate or useful simulations of intelligent behavior.
The Androids here are based on the Model of Hierarchical Complexity. The Model of Hierarchical Complexity is a general behavioral-developmental theory that applies to behavior of all animals, including humans and computer based models. That Model is consistent with evolution. The Orders of Hierarchical Complexity are mathematical model that may be applied to account for how organisms and groups of organisms behave. At present, the artificial neural networks operate at Stage 3 (Circular Sensory-Motor Stage) characteristic of insects. We have created a higher order neural network that thinks, perceives, understands, predicts and manipulates better than insects. Our goal is to develop higher order stacked neural network that emulate the hierarchical complexity of the smartest people as measured by their behavioral-developmental stage. Such androids should even be able to design future androids smarter than any human.|