Artificial Intelligence

The thought of machine-based intelligence to learn and decide upon given input, seems almost supernatural.  Artificial intelligence has been a vision and dream of thousands if not millions of people throughout the years.  There are numerous counts of artificial intelligence (AI) being researched, studied, and implemented over the course of generations.  Even though, the start of artificial intelligence was purely imaginary in theater, films, and movies, it gave artificial intelligence a bad impression from the horrible stories portrayed.

Although, the possibilities of utilizing artificial intelligence within robotics is a viable option and it is even being tested in science research labs across the world today.  There are even far more opportunities available when utilizing artificial intelligence separate of robotics, such as the well-known SIRI, newly created self-driving cars, and visual recognition.  As mentioned by Joachim Buhmann (2008) in a book section titled “Visual Recognition: How can we learn complex models?”, within the book Security Informatics and Terrorism: Patrolling the Web: Social and Technical Problems of Detecting and Controlling Terrorists’ Use of the World Wide Web:

Therefore, learning of these image categories should not be hand tailored by humans, but the statistics should speak for themselves and the algorithm should find a representation which actually warrants the label given to that image. (p. 166)

In the given study by Buhmann, there are three structural properties utilized for an artificial intelligence model to make various informed observations and decisions about the given input.  The first of the three is called decoupling, which is the process of breaking down the image into smaller portions.  The second is modularity, which is a method of separating and organizing the group of objects, such as pixels.  Third and final structural property is hierarchies, as the hierarchical dependencies can efficiently be utilized to enable an efficient search procedure.

For the given artificial intelligence study, the system is able to generate reasoning and learning with compositionality.  This compositionality can be viewed as a type of strategy, that allows inferring pieces which are exceedingly associated with the entity’s uniqueness and to translate the appearance variables by the relations amongst the pieces.  As described by Buhmann (2008):

Composition systems use a small number of generic parts as their key design concept. These parts are pieces in your data, which occur sufficiently often to code them as separate entities. More complex representations are then assembled by combining these parts to new generic features. This assembly is called a composition. (p. 170)

Thus, the generation procedure of reasoning and learning is a combination of comparing variables and the process of elimination.

Whether the results of the analysis toward the theory for reasoning was Boolean, Bayesian or both, concludes according to the study.  Considering, the basis of the artificial intelligence model consists of the Bayesian network, that couples the compositions, shape, and image categorization.  Thus, this development allows the system to filter out the relevant compositions that are available within the larger set of candidates.  The most logical next step is filtering out all relevant combinations of combinations and to utilize them for the decision-making process.  Training these composition systems is a quite delicate process considering, it is required to have sufficient regularization at each level of learning.

According to the research findings available, there are numerous computations that would not be appropriate for the artificial intelligence system within the reading.  Considering, the artificial intelligence system described within the case study utilizes observations and decision-making on images inputted, then any type of computation that does not provide an image would be inappropriate.  As an example, requesting the artificial intelligence system to decide whether a business should invest in one specific business proposition or if it should decide on another.  As Trapp discusses, “Indeed, the potential of AI to unlock the secrets in the ever-increasing amounts of data being collected is helping to transform a particular and important part of the prediction business - forecasting and planning.” (Trapp, 2018).  Therefore, this specific artificial intelligence system would not be helpful in a forecasting situation.  Although, the artificial intelligence system described is able to recognize patterns, which allows for backpropagation. 

Reasoning backpropagation is an essential component within artificial intelligence systems as it allows the system to quickly calculate the variables to make a decision within a timely manner.  Just as the name implies, a backward pass for an error to each and every internal node within the neural network.  This allows the system to utilize calculated weight gradients for that specific node.  Thus, backpropagation networks are essential to deal with the various types of available data.

An artificial intelligence system may seem to be a scary thought and almost completely imaginary, but they are already in existence and seem less frightening first hand.  Each system is taught how to handle specific observations and decisions, but the likelihood of one handling it all is still a bit out of reach.  Implementing artificial intelligence in an organization is not as farfetched of an idea as it used to be.  The benefits of the implementation within an organization are far greater than the risks.

References

Buhmann, J. M. (2008). Visual Recognition: How can we learn complez models? In B. K. Shapira, Security Informatics and Terrorism: Patrolling the Web: Social and Technical Problems of Detecting and Controlling Terrosits' Use of the World Wide Web (pp. 166-). Zurich: IOS Press.

Trapp, R. (2018, June 14). How AI Can Help Leaders Make Better Decisions. Retrieved from Forbes: https://www.forbes.com/sites/rogertrapp/2018/06/14/how-ai-can-help-leaders-make-better-decisions/#76b260e35e2f