W3 – Automatic for the People
Week 3 of Automatic for the People was like week 2, with our focus being to generate a film scene and represent a certain issue and genre. My group decided to experiment further with the scene we had scripted with ChatGPT last week. After having a discussion with Joel, we chose to prompt AI with different genres for the same script. For example, we asked ChatGPT to turn the script into a musical with stage directions included. We found that ChatGPT did not change the script no matter the genre… other than the musical because it created a song more so than a dialogue script. Our most liked script was when we prompted the AI to make a kids cartoon with our synopsis. After further prompting it to include soundtrack and sound effects, we all agreed that we could visualise the scene well and the script made sense which was a first. I still believe that there is always more to be added by the human brain. AI is exceptionally quick and can generate facts, however, it is basic and lacks depth.
This week I have seen videos on TikTok of people using an AI filter to determine what their baby/child would look like with someone else. You pick two portrait photos, and the filter produces an image of a child. Of course, people have multiple children and they can all look different, however I was so interesting to see people put in celebrity images and the child would look exactly like one of their children. I think it is an interesting filter but I think it is just for fun and cant be relied on to determine your babies appearance.
Ted Chiang’s article ‘ChatGPT is a blurry jpeg of the Web’ speaks of the quality of ChatGPT, explaining that it is not only ‘blurry’, but is also like an ‘unreliable photocopier’ (Chiang, 2023). Chiang’s article is great in explaining why we should not rely on ChatGPT and that even thought it seems very advanced, it does still have a lot of development ahead.
“ChatGPT is a Blurry Jpeg of the Web” – Ted Chiang, February 9 2023
LINK TO SKETCH