a man sitting on a couch looking at something
a man sitting on a coach looking at something
Every half an hour for the last week, a green light has lit up on my laptop informing me that it’s currently taking a photo. Each one of these photos was then translated to text by a machine learning algorithm. These snippets of text were then compiled into a log, an automated AI Diary of my week. Over the course of this week, what I originally believed would be a playful and interesting experiment on self-documentation, left me feeling exposed, uncomfortable and dehumanised. So this month I want to tell you about my week of self-surveillance, where it took me, and what I learned.
The project started last Monday when I had nothing to write about in my newsletter. I had been planning to write about the concept of hyperreality and my recent fascination with the impact media can have when it refuses to be easily classified as fiction or non-fiction. Things like Nathan Fielder’s “The Rehearsal”, Werner Herzog’s “Family Romance, LLC” and Ross Sunderland’s “Imaginary Advice”. However, I soon realised I didn’t really have a point to make yet and was just about to ramble.
So with that idea shelved, I set about trying to find a mini project that would allow me to learn a new skill. I settled on expanding my knowledge of machine learning and how to run my own models locally using python. By the end of the day, I had managed to set up a small script that took photos every 30 mins when my laptop was open and awake, and would then use an AI model (vit-gpt2-image-captioning) to caption these photos. Thus, hopefully, creating a sort of automated diary of my week.
The first two days of my self-imposed surveillance were novel and amusing. It was interesting to see the limitations of the model, the way it repeated itself, and it’s inability to be self-referential. Hours spent working at my computer were reduced to ‘a man sitting on a coach looking at something’. The computer struggled to realise ‘it’ was the something I was looking at (working on). Equally, the diary continually stated I was sitting on a sofa (coach), as my body obscured the view of the chair I was actually seated on. So rather than assuming I was levitating, the AI must have seen my bus seat like sofa behind me and decided I must be sitting on that. However, the charm soon wore off as my life was reduced to the simplest 32-character description.
After the first day of the project, I committed myself to uploading and sharing the full log. It felt as though the only honest way to engage with this idea of self-surveillance was to do it with the knowledge that the output would be public. Hence, I tried my hardest in the first few days of the project to show the camera, and thus the humans who viewed the project, that my life was interesting, healthy and full. I began to engage in the fraudulently authentic behaviour I had previously scoffed at when my friends started using BeReal. Tomas Flight talks about a similar concept in his great video essay on ‘The Rehearsal’, where he discusses the way in which reality TV show participants struggle to act like their authentic selves but instead fall into the role of entertaining TV show participants. This is due in part to how they are edited by the producers, but also because their only exposure and education on how people act on TV are from other reality TV programs (which in turn were edited). So much like a reality TV show contestant or someone living for the ‘Gram’, I tried to modify my behaviour for my audience. However, my editor was not so kind and had no eye for entertainment, so my new habit of reading rather than consuming YouTube, was translated to ‘a man sitting on a couch with a remote in his hand’ and my morning exercise routine to ‘a person standing in a living room with a couch’.
As the week progressed, I spent time creating a website to show this AI Diary, which required me to keep looking at this log of my seemingly dreary life. I now felt awkward and uncomfortable whenever the green light came on, indicating another photo and another unflattering caption. I refused to just shut my laptop lid, that felt like I was cheating or defiling my own ‘scientific’ experiment. So instead I made an effort to not be in my room for the occasional 30-min green flash, each time assuring myself I had legitimate reasons to do so and I jumped at any excuse to leave my flat entirely.
On Friday night I had no plans, but I was excited to make a nice dinner, put some candles on and get comfy for a long film (Raging Bull, for those interested). This excitement faltered under the camera’s eye, what would people think of my lack of Friday night plans? I was put in mind of Jennifer Ringley, who live-streamed her whole life from 1996 to 2003 in a project called JenniCam. I had referenced her in a project I did at university about creating a more ‘authentic’ form of social media. One where nothing was filtered, and in so doing, people would feel more empathetic and embracing of the realities of human life. The inspiration for the project came after hearing that someone had emailed Jennifer after seeing her doing laundry on a Saturday night. The emailer wanted to tell her how comforting it had been to know that they weren’t the only one without fun Saturday plans. However, as the one doing the metaphorical laundry, with the sounds of busy Berlin out my window, I just felt exposed and embarrassed. As the rest of the experiment continued, I found more ways to lie by omission to my own project. Choosing to read at a cafe on Sunday, watching TV on my phone rather than my laptop, only exposing parts of my life I felt appropriate.
As I sit here on Monday night, the green light just blinked at me, writing this, I am able to find some solace in the fact that the amount of data generated and the simplistic nature of the output, will never really fully expose me or my life. In fact, I am sure that this newsletter is an order of magnitude more exposing of my neuroses than any AI could be. However, now that the project is concluding, I am left with a few takeaway thoughts.
Firstly, throughout the project, I honestly felt surveilled. I felt this even though I was aware of what was happening, controlled the instruments of the surveillance and had the power over the output’s publication. So why don’t I feel this same uneasiness when using the internet? Google’s own spying operation is definitely more advanced and all-seeing than my own model (which somehow believes my jacket is a dog, my jumper a cat and my reading light a fireplace). Would seeing this log of our activities online make us value our privacy more, or did I only feel so uncomfortable because I know my friends and family (Hi Mum) will have access to this log of my life?
Secondly, and this may sound contrary to my first thought, I was shocked at how little beauty or nuance the AI was able to capture about my life. I was an object, in a room, doing a thing. I was ‘a man sitting in a living room looking out the window’, not a person trying to get away from his computer for just a second, so they could ponder if this project was a good idea. I felt dehumanised by the green light, in a way I think we all are by technology. Reduced to moving parts, understood only enough to be useful to the technologies’ own needs.
During this experiment I was reading Douglas Ruskoff’s ‘Team Human‘ and in it, he speaks to the limitations of Computers, and their inherent anti-humanness. When talking about a computer’s ability to understand the world, compared to humans, he says: ‘Computers work closer to the way primitive, reptile brains do. They train on the foreground, fast-moving objects, and surface perceptions. There’s a fly; eat it. The human brain, with its additional lobes, can also reflect on the greater spatial, temporal, and logical contexts of any particular event. How did that fly get in the room if the windows have been closed?‘
On that note, that’s all from me this month, thanks for reading and I hope you have a great week,
‘a man sitting in a living room looking at his reflection in the mirror’ (Fred)
Read the logs here: https://ai-diary.mildlyupset.com/
-
My Website: https://www.fredwordie.com
My Book: https://www.bigdatagirl.com
My VC: https://ventually.xyz