Different Approaches to Artificial Intelligence Engineering

On the other hand, one major disadvantage of Alexa is privacy. (“Amazon Alexa,” 2018) Others can hear your conversations with Alexa, and whoever is present in the home can hear in on private conversations. Alexa also requires you to give your current location, which can be accessed by other servers or even hackers. (Phillips, 2018) Another disadvantage of a narrow AI is that the intelligence cannot work outside of what they were programmed for. Therefore, an AI cannot make connections to certain situations as well as humans could. (Legg & Hutter, 2005) Lastly, humans can become too dependent on AI and can risk losing their mental capacities.

An example of Artificial General Intelligence is a robot named Andrew from the movie “Bicentennial Man.” (Columbus et al., 1999) Andrew is comparatively as intelligent as the human brain and has human-like characteristics which allow him to experience human emotions and creative thought. (Columbus et al., 1999) Unlike Artificial narrow intelligence, the robot learns to improve itself to perform various tasks such as visual perception, decision making and speech recognition. (Columbus et al., 1999; Goertzel & Pennachin, 2007; Huang, 2017) Andrew would be identified as a learning agent because he learns from experiences and is capable of learning from mistakes. (Goertzel & Pennachin, 2007; Russell & Norvig, 2009)

Furthermore, he is a learning agent because he can learn and solve new problems that he did not know of when he was created. (Columbus et al., 1999; Goertzel & Pennachin, 2007; Russell & Norvig, 2009) Andrew would not be considered a simple reflex agent because Andrew has a deep understanding of the world and is able to think through situations, rather than give an automatic response. (Columbus et al., 1999; Russell & Norvig, 2009) He is also more advanced than a goal based agent, because his actions do not depend on goals to be reached. Rather, he is able to alter his actions to his specific emotional needs and the emotional needs of those surrounding him. (Columbus et al, 1999; Russel & Norvig, 2009) Lastly, Andrew would also not be considered a utility-based agent because although he is not trying to optimize his goals and desires, rather he behaves just like a human being. He is capable of forming goals, and learning from them if mistakes are made. (Columbus et al., 1999; Russell & Norvig, 2009)

Artificial general intelligence is an intelligence that is as intelligent as an average human. Robots like Andrew would have a low error rate compared to humans, if coded properly. He has incredible speed, accuracy, and precision that allows the intelligence to replace humans in tedious tasks and in difficult places to work. (Columbus et al., 1999) Additionally, they will not be affected by hostile environments and can endure problems that would injure or kill human beings. Therefore, this would allow them to complete dangerous tasks, such as mining and digging feels that would otherwise cause problems to the health of human beings. Furthermore, robots do not need to sleep, eat, or take breaks as they do not get tired or bored. Thus, would be very productive in work situations. (Columbus et al., 199) Lastly, another great advantage of AGI is that they can assess people for medical purposes such as health risks and emotional state. In the near future, robotic surgery can achieve precisions that humans cannot. (Columbus et al., 1999)
 


While there are numerous advantages of AGI, there are also various disadvantages. There are many ethical issues concerning AGI. Specifically, an AGI could develop hostile attitudes towards humans and in the worst case, could lead to the extinction of the human race. Another disadvantage is that AGI can cost a lot of money and time to build, and repair. An important disadvantage of an AGI is that with their human-like characteristics, they can replace our jobs leading to severe unemployment. Lastly, AGI machines can easily lead to destruction, if put in the wrong hands a robot like Andrew can supersede humans, enslaving us. (Columbus et al., 1999)

There are two distinct artificial intelligence cognitive engineering approaches, constructionism, and constructivism. (Thórisson, 2009) Constructionism is manually pre-programming an intelligence with algorithms and procedures. Everything the intelligence needs to know will be fed to it during its creation. (Thorisson, 2009) Alexa is an example of the constructionist approach. This is because Alexa is a pre-programmed virtual assistant that is fed with information during her creation. She does not gain knowledge from learning, but rather her intelligence comes from what was programmed. (Thorisson, 2009) On the other hand, constructivism is when an agent has to learn from mistakes, and the knowledge grows as the agent gains new experiences. This is seen in the movie “The Bicentennial Man” in the robot Andrew. Andrew is given some information at his creation, and over time that knowledge grows as he learns from his mistakes and experiences. (Columbus et al., 1999; Thorisson, 2009)