Solid State

Reflections on Workshops

Week 3 Web Scraping

This week’s workshop on HTML and web scraping provided valuable insights into web development and data collecting techniques. The session began with reinforcing foundational skills like structuring a webpage using tags and attributes. The tutor introduced several useful browser extensions and applications, such as "Web Scraper" and "Outwit Hub," which simplify the process of selecting elements for scraping. These tools really enhanced my efficiency and understanding. I gained many useful skills in this two hours, and I plan to further explore web scraping, possibly by working on personal projects that involve data collection and analysis.

Week 5 Data Visualisation

In this week's workshop, first of all, we did some critical thinking about the dataset we collected and generated earlier, and the following is a reflection on this part: my group collected user reviews and ratings of an AI chatbot's software after the class as a way to understand how AI can shape our understanding and perceptions of intimate relationships. After mastering the web data crawling skills from the previous workshop, collecting this data is not a difficult process, but how to create a perfect cut research meaningful dataset is a serious issue, for example, how to define the type of comments collected is the first question that comes to my mind, whether to categorise them according to the emotion level or according to the specific content, because this directly affects the the research significance of the whole dataset; the second problem was how I could make connections between different variables to make the results more intuitive and effective after data visualisation.

After that, Holly explained data visualisation through excel and allowed us to create tables and manipulate them ourselves, the only problem I had was that I wasn't sure about the last part of the pivot table, but I solved it through self-study after the class. I was shocked by the efficiency and convenience of data visualisation in dealing with huge data. Compared to my own dataset, the even bigger dataset of the UK Resident Survey shown by Holly in the workshop was even bigger and more complex, so I can imagine that it would be impossible to deal with these data and get valuable survey results without the help of data visualisation, and at the same time, I have also learnt a lot from the codebook in the class file. codebook in the classroom document learnt a lot about coding and defining variables.

Week 7 Machine Learning

This has to be the most interesting workshop I've been to so far, in this week's workshop we were introduced to machine learning and training computers to recognise the content of images through the machine learning web site. At first I was excited about what this website could do, as if I had some kind of black magic, but as Holly showed us how the model made mistakes and let us try to mess up our own trained model, we gradually learnt that the model works by recognising the pixels of the image it captures and comparing them to the similarity of the images that we uploaded during the training process. For example, when I uploaded a set of gender-identifying image clips, which included thousands of images, my model captured different angles of me as different genders after the model was trained. This shows that despite the large amount of training material that has been input, the code logic of the model leads to limitations in its functionality.

This question also got me thinking, machine learning is already very common in our daily life, such as face recognition in mobile phones, or artificial intelligence everywhere, and furthermore, personalised content recommendation, image and voice recognition, etc. are all advanced variations of machine learning, so at the root of the problem, the application of machine learning is to minimise human involvement while accurately completing a simple single mechanical task. The application of machine learning is designed to minimise human involvement while accurately completing simple, single-mechanism tasks, but is our privacy or critical information really safe in such an environment, which brings us to the key question of how to regulate and control AI.

Learn more