MIS 2101.730 – Courtney Minich – Spring 2018

Weekly Post #6

Leave your response as a comment on this post by the beginning of class on March 1, 2018. Your response only needs to be about three or four sentences. These weekly questions should reinforce class discussions, readings and show some insight.  

Artificial Intelligence is rapidly advancing in today’s society. It’s presence is not only in real life but also in the entertainment industry. A Netflix series, Black Mirror, discusses artificial intelligence and machine learning in various episodes. Watch the Black Mirror episode titled “Be Right Back” [Season 2, Episode 1] and discuss the different areas in which machine learning [https://en.wikipedia.org/wiki/Machine_learning] is present throughout the episode. 


22 Responses to Weekly Post #6

  • Machine learning was present to the nth degree. The entire premise of creating a digital version of a passed love one, is but a small part of the global artificial intelligence world.

    We see several types of AI employed, but I will show two.

    Decision tree learning – We take a normal decision process and using a persons historical choices predict the choice pattern. In this case, it was the words Ash used when speaking. Ash was sarcastic and the AI recognized that and used that in many responses.

    Deep learning – Using voice profiling, coupled with emotional interpretation of writing styles and voice modulation, added to political opinions and even sarcastic tendencies, AI can learn just as a person learns from another. Sarcasm being the largest obstacle to overcome. To a computer, a True or False statement is a 1 or a 0, something or nothing, yes or no. Sarcasm introduces a conflicting statement of truth. A lot of humans struggle with sarcasm, but with computers it is far harder. In programming, all statements are valid arguments, if not, the program will not execute. However, deep learning allows complex computations of validity, allowing contradictory statements to exist.

    I could go on and on, outlining almost every example from the wiki entry.

    The real pain was watching the women’s heart break every time the AI encountered an emotional/reactionary situation. Ash couldn’t react on an emotional level -when he found himself on the other end of a heated argument- through primitive brain function, that cerebral inhibition just wasn’t there. He could also not stop himself from jumping based on fear, another reptilian brain function. Pain centers were not there after cutting himself on the glass.

    As a complete aside, that is some of the best acting I have seen in a while. I literally felt that poor women’s heart break every time they interacted.

  • When Martha is first talking to Ash via chat you could see the decision tree learning process. It progressively learns how to respond to her basic interactions over chat and then over the phone. The downfall was when Ash the robot came to life and was unable to reproduce the nuances that a real person would have. Robot Ash was based off of real Ash’s social media persona (a lot of unstructured data), which doesn’t capture all of a human life. Robot Ash can’t respond to emotions like a real human being. When Martha gets upset at robot Ash, he asks did Ash do xyz, to figure out what type of response he should do next.

  • I think machine learning happened with every interaction with Ash. When the account was set up and The initial search for public information was done to set a default personality that started emailing the woman, Each time she corresponded with him via email he learned and adapted, each time they talked on the phone ash Learned and adapted, also all of the physical, verbal and nonverbal interactions between Ashley and the woman was a learning environment which ultimately stopped her from having him jump off the cliff.

  • The first machine learning is involved is in the feeding of social media input to attempt to create a programmatic voice for him. This would fall under “supervised” learning, as the machine is being fed input under supervision.
    The second way I see machine learning being involved is the initial texting and talking phase. This could be using reinforcement learning where the feedback from the user is further used as training data to improve the AI.
    In addition, I would expect the avatar creation to also be done via machine learning based on say, Instagram posts, given to the AI as a supervised learning training set.
    Finally, the scene at the end, where she tells the AI to react and show fear, and it starts to, is also an example of reinforcement learning, whereby she is providing training data as reinforcement to improve the AI.

  • Machine learning is present throughout the episode. To begin with, “robot Ash” has to learn how Ash would respond or react to things from his social media activity, emails, and text messages. It gleans everything that it can from these sources, but still does not know how Ash would respond or react to Martha in emotionally charged situations. I think that this is an example of supervised learning, where the machine is given inputs and needs to create outputs. Martha’s responses to what “robot Ash” did or said were reinforcement learning that told it if it was acting like Ash or not, and if it needed to change. She also gave the machine more inputs when she talked about Ash and what he used to do. She taught it to close its eyes when sleeping and tried to teach it to breath naturally. She also taught it that a normal human being would be afraid of death. Despite all of its capacity for machine learning, Martha is not satisfied in the end because it is still not Ash. Very interesting premise! Reminded me of the movie Transcendence [https://en.wikipedia.org/wiki/Transcendence_(2014_film)].

  • Much like when you play sports sims, the game learns your favorite locations you like to shoot from, and which players you like to use to shoot certain kinds of shots or passes you like to throw or which side of the field you like to run to, artificial intelligence is present. It learns which plays you like to run in certain situations, therefore the more you play it learns your habits, strengths and weakness, and how to exploit them. Yes, Machine learning! Which brings me to Black Mirror episode titled “Be Right Back” [Season 2, Episode 1] and synthetic Ash began learning about the real ash by scouring the web, especially Ash’s social media pages, text messages, voice mails, and emails to learn the basics of who Ash was. As the synthetic Ash, and Martha began to speak to it and correct it to what the real Ash might say or do, she also inputs photos into it he, the AI learns what to do and or say what the real Ash would do. I think when the synthetic Ash took on human form, Martha’s expectations rose to a point that synthetic Ash couldn’t meet Martha became upset with it. Also, I think there are some emotions that it artificial intelligence is not capable of, that is where her limits were pushed.

  • Machine learning obviously the backbone of this episode. it’s present on every level and I think two of the biggest pieces of machine learning we see are decision tree learning and deep learning. Decision tree learning is present when Martha is talking to the AI version of Ash through the chat system. The AI uses data from social media to predict and decide how it should respond. Deep learning is present when the AI version of Ash gets created and actually takes physical form. At this point the AI is processing speech, audio and learning to act as if it were a normal human being.

  • Ash’s program used machine learning just to come into existence. Martha’s friend (?), allowed the program to have access to Ash’s social media to become “program Ash”. From there Martha fed the program videos of ash that it then created a voice with. Throughout every exchange that Martha had with program Ash, the software was learning and evolving to become more like real Ash. It is unclear (at least to me, I might have missed it) how Android Ash determined what the look of real Ash was, but we can assume that this was also learned by the program through pictures and videos. After android Ash came into existence, he continued to learn about the ways in which real ash would behave.

  • Human simulation via machine learning is seen through out the whole episode of “Black Mirror”. Robot Ash learns how to respond to Martha using Ash’s social media, emails and charts. It mimics Ash’s voice and the way he speaks. This’s an example of supervised learning. Unfortunately, robot Ash doesn’t know how to respond emotionally to Martha’s feelings. When Martha makes robot Ash to show fear, it does so, and that’s an example of reinforcement learning. In general, the whole episode is kind of creepy. The AI is developing so fast and definitely there are lots of advantages of it. But on the other hand, it frightens me how our future will look like with robots and human simulation software.

  • While I had seen this episode of Black Mirror previously, it was just as good watching a second time. The first application of machine learning in the episode was a program gathering someones “digital life”. By going through all of a persons online communications such as social media and anything previously recorded they were able to create the basis of that person. Of course for this to work properly there must be enough data available. Later, the “machine” or imitation person learns further from actual interaction. Once the AI system (or person in this case) has a strong enough basis in the history and personality of the person its imitating, the AI only needs to continue learning and improving to be complete. Of course the show brings up other issues such as whether an AI can truly be “human”. In regard to the machine learning, deep learning is certainly present in the AIs (or Ashs) ability to mimic the human he is replacing. According to the wiki on deep learning, in the fields it has been applied “they have produced results comparable to and in some cases superior to human experts”.

  • The whole episode of the black mirror is based on the machine learning. Creating a using technology person who has been dead is just remarkable. The first machine learning process was to create an online profile of Ash’s using all of the information available online about him such social media, recordings, and videos. After that Mar uploads all the videos into the program she had of Ash’s and it evolves into a person who can talk on a phone. After that its taken to next level she orders an actual robotic physical body that looks like Ash and this robotic Ash uses every bit of information installed in the program to create a personality that is similar to how Ash was in his actual lifetime. Even though it wasn’t a perfect reflection of Ash, it was pretty good. It was my first time watching this episode it was very very interesting as to how is our world going to change in the future.

  • I agree that machine learning is present throughout this whole episode of Black Mirror. We saw “decision tree” while robot Ash was interacting with Martha. All of the information was gathered through social media and pictures or videos, which lead to his responses, but he could not form his own genuine reactions. We saw that he could not when Martha got mad and wanted him to hit her, and when she asked him to jump over the cliff. Throughout the episode I felt that it was a little creepy having a robot of your dead husband, but if available in the future I can understand a person being so heartbroken over the death of a loved one, that they would resort to such a thing.

  • In this episode of Black Mirror we see the robot, Ash, use machine learning. The robot uses machine learning by constantly adapting the way it responds to Martha; it is learning the way Martha wants it to respond. The robot also uses deep learning to respond the way Ash would of. At first it doesn’t quite mimic the way Ash would, but at the end when Martha and the robot are standing on the cliff, the robot reacts how Ash would; begging and pleading for its life. On a side note, another amazing show about AI’s using machine learning is WestWorld. I highly recommend it.

  • The concept of machine learning was present throughout the entire episode. The robot “comes to life” through supervised learning, using decision tree methods based on information gathered about the deceased Ash. The robot is able to take on many characteristics of the man, but not all, which is evidenced as the episode moves along. Each interaction with the robot served to “teach” it something about how it should act or respond in various situations. I found the whole thing a bit disturbing, especially when thinking about how this is becoming a reality in our world.

  • In the episode of black mirrior, we see the recreation of Ash learn more and more about “himself”. It begins as just text and eventually becomes a full on android. The machine learns in many ways throughout the episode. It learns though photos, videos, and chatting with Martha. The machine learns a lot, but this episode shows that it would be hard to perfect, with the machine not choosing the correct action at times. They learn based off of information, and the information does not always show everything, especially for a human.

  • After everyone has watched the episode the answers will be similar.
    After Sara signs her up for the software the software uses public information and Ash’s online activity to create a baseline for queries(questions) and responses. Once that is done Once Martha wanted more she uploaded private information via video to give vocal expressions to the software. The machine then learned to speak as Ash after analyzing the data. As the computer learned Martha’s personality it was able to analyze wants thus asking if she would like to hear a way to see him. The machine was then given physical form via the unstructured data it received. It was able to express via the videos previously given to it. As the machine functioned and interacted with Martha it learned from its experiences. It still lacked the human quality of consciousness and emotion. The robot was just showing what was previously recorded unable to think for itself, it served its purpose even referring to Martha as the administrator because it learned based on what she provided. All in all the concept was very unique and also a bit weird. We saw the machine grow and begin to act as Ash even getting wittier with the responses as he interacted with Martha.

  • The sings of machine learning regarding Ash begin when he begins to “message” Martha and very early on in the conversation he says something that brings a massive smile to her face. Her response is, that’s just like something he would say. At this stage the AI only has Ash’s public information to go on. Later on based on videos uploaded by Martha, the AI can actually replicate Ash’s voice and as well as tone. Later Martha explains to the AI that Ash uses a particular term for a “freak out” The AI within seconds analyzes this new information and apply’s it in the right context. The AI uses Martha’s “orders” instantaneously, really to a fault, as Ash’s true personality would not allow him act so complacent. I can only speak for myself, but I thought they did such a great job of making the AI system seam human that I actually found myself feeling sympathetic towards the “character”. This is the first episode of black mirror I have seen, but I am certain it will not be the last

  • Machine learning was very much present throughout the duration of the episode. The program “Ash” would learn how to mimic the way he would speak by going through any of Ash’s public social media accounts, and then it would learn to vocally speak by taking the sounds of the real Ash’s voice through video’s that were uploaded. The robot-like Ash would learn from experiences as time went on, but still could not form human reactions, further showing that these programs and machines can not replace the real thing.

  • It is somewhat scary to think artificial intelligence and machine learning could be as advanced as it was in the episode. Creating a response messages and caller is one thing. Activating an artificial life form of a dead loved one is something completely different. “Ash” was able to analyze data that had been posted by Ash in order to recreate similar responses. To make this more realistic personal messages, videos, voice recording, and Martha’s memories were also analyzed. However, every human has unique thoughts and memories that they can’t/don’t share, because of this “Ash” was never really able to be Ash. As smart and programmable as “Ash” was, he wasn’t human, and no amount of machine learning, at least not in this episode, was going to change that.

  • We see machine learning at the very beginning of the show and pretty much through out the rest of it. We see Ash uploading texts and photos on what appears to be some kind of social network. I gather he was feeding the “machine” information about himself which was then later turned into a superficial being of Ash after he died. As the show goes on, Martha starts using an app like thing to connect and communicate with Ash. She teaches him some new things and even shows him the environment around using the camera on her smart phone further enhancing his intelligence. The superficial Ash is countinuing to learn even after he has been gone.
    After Martha received the robot that mimics Ash, she continues to teach him more human like behaviors. One very interesting part was when Martha got upset at the robot Ash for not exactly responding like the real Ash would’ve. I think this shows us that AI can not be completely human in as far as expressing emotions. But at the moment, it can process a lot of thinks (logic) which us humans sometimes tend to fall short on. Interesting show. I am definitely going to watch all of it.

  • Machine learning was noted throughout the episode. Ash started by learning through email, and then expanded through watching videos of the real Ash. He was literally mimicking what the real Ash had left recorded. I found this episode sad, as she was looking to replace her fiancee through artificial intelligence. After she taught him what she could, she realized that he was still a sad copy of the original. Machine learning in this aspect, has limits as to how close it could get.

  • I must say, I will begin to watch Black Mirror. I did not like the concept of this episode, just because I lost my mother at a young age, and I do not know if I could reintroduce her back into my life; I don’t know how that sounds to people, but people learn to cope. Additionally, I believe there is hidden symbolism in the episode. As Americans, we have an everlasting desire for something more, especially technology. We always want the newest iPhone, the newest Snapchat update (not so much right now) and the newest technology (VR) in general. Concepts like this will keep us wanting more, which can be taken too far, like bringing someone back to life in the form of a “human.” All in all, technology and A.I. will continue to develop and change beyond our wildest dreams.

Leave a Reply

Recent Comments