WEEK5: Winnie Soon

Updated: Jan 16, 2020

At first, when I listened to Winnie Soon's lecture, I didn't have any interest. Maybe its because of the artwork is quite dull to me.

However, when the lecture was talking about more detail about her artwork's process, it is getting more interesting. I like the way that she was talking about Machine Learning. In my opinion, we know the data is easy to capture on the internet, but the process is still a mystery for all of us which is connected to" Algorithm" Gillespie -2014.

Three-point in ML.

1.Input training data

2.Training Modle

3.Output prediction

The essential part is the input data, with such a massive database, how can we select the right data to the Artificial Intelligent to learn. This is a fascinating point, and she has to select the right data to AI by herself in the present time. I believe that some people create a program to let AI select the right data, but who knows the different definition about right or wrong between human and AI.

Long Short Term Memory in ML.

For Winnie Soon, the LSTM is one of the exciting parts in her research study. For us as a human, we have LSTM to remember the different memory which is vital to us or not. It is a provoking technology to use this in the AI. This gives AI doesn't have to keep all the data in the memory but also can collect the data its need, which able to clean the data by themselves.

Data distribution.

For the development in the ML, the DATA is the most vital part of it. However, most of the data keep in the particular head of the tech company such as facebook or google etc. This means they have more advantage in the development of the ML/AI. It is crucial to think about the data distribution, obviously more data which means more power nowadays. But is that means the public doesn't deserve to have the data resources develope more dipper in this fascinating field? Also, we have to think about the data right and privacy on the internet, what if there is some data that we don't want to leak out but still being used by the research.

ML in Chinese poetry

One of Soon's work is to develop a program which AI will create Chinese poetry, and she used the Chinese social media "Weibo" as the database to collect the Chinese words data. We can see from her process book, which shows at the beginning, the AI create the meanless sentences; some of the words are even not Chinese. As the AI gets more data and practice, the Chinese poetry that AI creates is getting more accurate. As an example, we can clearly see the ML process in this case and think about what it means if the AI can create poetry.

Censorship in China

ML can be used positively but also negatively. The example that Soon's points out is the internet censorship in China. She shows that in the 4th June all the information relates to the Tiananmen Square protest will be blocked more seriously. Chinese people cannot even mention about "Today" on their social media. However, people still trying to use a variety of different code to say about the Tiananmen Square protest on 4th June 1989 but again being blocked by Chinese censorship.


Recent Posts

See All