Tag Archives: Deep Neural Networks

Deep thinking about Artificial Intelligence in the movie “Her”

Recently, I’ve been recommending one of my friends to watch the movie “Her“. The movie “Her” is one of my favorite movies. This is due to two things: it can be interpreted in many different ways and also because the Artificial Intelligence (AI) fell in love with the main character, Theodore Twombly, which was very new and impactive idea to me. While thinking something interesting to post on the blog, I thought of some key words, such as brain, computer, memory, and intelligence.

There are a lot of AI movies; robot movies, which talk about the future world. At first, it started as pointing out people’s alienation as technology began to become increasingly highly-skilled. The imagination in these sub-genre of movies made the audience consider the possibilities of a world riddled with new technology and also the problems that followed as a result. However, our current world is already high-tech. Researchers and inventors are already making a lot of devices that were influenced from movies.

An article written by Vlad Sejnoha mentioned Deep Neural Networks (DNN) and the high accuracy of Samantha ( the invisible woman played AI role in the movie). A deep neural network (DNN) is an artificial neural network with multiple hidden layers of units between the input and output layers.  DNNs depend on ‘learning from examples’, the networks are classified with the labeled training exemplars and learn relationships between the input datas and the desired classification. Like most of you, I have not heard of DNNs as well. However, this concept is already into our thinking naturally. According to the article, DNN’s topology mimics brain structures and it is easy to understand AI by interpreting its DNN. In the movie, the AI Samantha seemed different than most computer in terms of her processing. While a computer is processed by a serial processing system, AI is processed by a parallel processing system just like our brain. AI combines symbolic processing and a machine learning system to help it adapt to unexpected situations. AI can multi-task, tasks like speech recognition, natural language understanding, speech generation, dialog, reasoning, planning, and can even learn new things. The following is an example of a neural network pattern shown in Samantha’s speech in the movie. “Sorry, nothing’s available until 9pm. Would you like another Italian restaurant in the area at about 6:30pm?” How can an AI suggest a restaurant and take into account the time, location, availability of the attendees and the type of restaurant? This is what neural network matching does; it learns the connections of each possible chance and the desired next step from trial and error. I am not going to talk about DNN deeply, but anyone who wants to learn more about it than speech recognition, can watch this easy youtube video. https://www.youtube.com/watch?v=M2IebCN9Ht4

AI interactions with humans is not a new concept, but this movie is new in that it’s based on emotional intelligence. In the movie “The Island”, the main characters are clones but have emotions and have the self-actualization that their existences are not real. “The Island” was released in 2005, and the movie “Her” was released in 2013. It’s been 8 years and a lot of things have changed. This movie is based on the emotion Samantha feels. She also gets confused about her feelings of love, worry, happiness, sadness, just like a human. This journal provides a good debatable issue, “Is it necessary for Artificial Intelligence to have emotions like a human?” I think it is unnecessary for AI to have emotions but they can be systemized to speak a right sentence in right situation in order to make a human to feel an emotion to them. To bring it in the real world, a current AI could be without emotions but if it could make its human user experience a plethora of emotions could be a more sophisticated machine. If you have a similar or different opinion, I want you to share with me and other readers.

  To close the post, in the future, I hope nowadays’ technique could deal with unstructured information such as pre-structuring of information sources, reasoning for not only superficial, but also introspection. In addition, as the author of this journal said, we should consider AI as not just an Artificial Intelligence, but also an Amplification Intelligence.

  Lastly, this youtube video shows ten high-tech movies so if you are interested in other good hi-tech movies, watch this.

Source: http://www.wired.com/2014/02/can-build-samantha-tells-us-future-ai/, wikipedia, youtube, http://www.imdb.com/title/tt1798709/?ref_=nv_sr_2