Over the past week, I have been conducting research on Generative AI through articles written about the topic by MIT. There were some articles about general information, many which addressed its uses in a business or workplace setting and the economy, and others which spoke on specific benefits and drawbacks that come with the use of Generative AI. The articles which interested me the most were ones about general knowledge about the topic and those that touched on the drawbacks, specifically the biases and inaccuracies, of the content.

https://unsplash.com/photos/a-black-and-white-photo-of-a-bunch-of-buttons-SUHcTWGuaUY

A common trend in all information I have read is that Generative AI is biased towards gender, race and political affiliations. This is because the information the AI is trained on is the knowledge on the internet which is full of accurate and incorrect information. Also, each time the site is used, the AI is learning from each interaction. Therefore, if someone is asking biased questions, then future results will be biased. I would like to gain a better understanding of the biases in AI and to better understand which areas the sites have more accurate knowledge.

The Plan

Photo of the ChatGPT Application icon
Photo of the Perplexity AI Application icon.

Using ChatGPT and Perplexity.ai, I am going to ask them to explain a movie to me. I am going to compare the information both sites provide and then watch the film myself to see how accurate their descriptions were. All the descriptions will be recorded with photos or videos (depending on how much information is provided) and will be posted on my blog. I will then write my own summary and compare which is most accurate.

Because I want to see not only how accurate the information is but also how biased, I have chosen a variety of films from different time periods, nations, and on created for different audiences.

Films I will be watching

  1. Gladiator (Oscar winning film 2000)
  2. Parasite (Oscar winning film from South Korea 2019)
  3. Unlocked (South Korean film not very well known)
  4. The Shawshank Redemption (1994)
  5. Inside out 2 (2024
  6. 3/10 to Yuma (2007)

Hopefully by doing this, I will be able to gain a better understanding of the depth of knowledge Generative AI has and can see how the information differentiates based on different nuisances.