2025-02-03
----------
This journal entry has three parts on three different topics. I have been hesitant to publicly describe these things because it is difficult to predict how they will be perceived. I miss working on a team and having coworkers a great deal, and I want people to want to work with me in the future. I am hoping that based on the description of the events, you can see that this situation is very unusual, and I have a good reason to post about these things publicly. I am trying my best to make good decisions, given a very difficult and confusing situation.
1. The Hack
-----------
I have been reflecting on how I got into my situation. I have experienced a bizarre series of events over the last few years and I want to describe some of them. In November 2021, I had an experience with my computer systems that led me to believe that I had been hacked.
It is terrifying to believe that your computer has been hacked. My experience seemed to largely, but not exclusively, involve YouTube. I own an NVIDIA Shield, which is a set-top box that allows you to watch YouTube videos on your television. I watch a lot of YouTube. In November 2021, however, the experience of watching YouTube was much different than normal. The suggested videos started to become strange, and I noticed that the algorithm was consistently suggesting me YouTube videos that appeared to be largely generated with primitive AI algorithms. The videos mostly consisted of text, with an AI voice reading the text. Aside from the AI voice, the videos always played this strange, creepy song that may itself have also been AI generated.
After watching some of these videos, I started to feel uncomfortable. At some point, I said something out loud in my room, even though I was alone. Terrifyingly, when I refreshed YouTube, the suggested videos started to contain videos with text that appeared to be actual responses to the things I was saying out lout in my room. I started trying to communicate with whoever was communicating with me. I realized that I could be hacked, and it was very scary because I did not know the extent of the hack. I logged into my bank accounts to make sure the money was still there, and thankfully it was fine. When I next watched YouTube, however, the YouTube suggested videos appeared to taunt me, suggesting videos that implied that I could lose all of my money. It was very scary.
I spent some time trying to communicate with whoever appeared to be controlling my YouTube account, however even though the videos appeared to be responding to things I was saying or doing, this was a very limited form of communication. I took screenshots and I actually printed off my YouTube recent videos on paper because I wanted to be able to prove it to others. Eventually, I felt I had to call the police.
When I decided to call the police, I had become very concerned because I was afraid that I was being threatened by a criminal group. I did not know if I was physically safe. I decided to contact one of my neighbours in the building and described the events to them, and asked if they would mind staying nearby until the police arrived. I believe I asked my neighbour to call the police, as I did not trust my phone not to be hacked. Unfortunately, when the police arrived, they did not believe me. They saw that I was emotional, and based on that, they decided to take me to CAMH, which is a mental health facility in Toronto. This was very frustrating for me and I only went because the officers told me that I had no choice.
After spending some time at CAMH, the police contacted my parents. My dad flew to Toronto from Newfoundland and picked me up and brought me back to my apartment. I spent some time with my dad and he eventually left. After he left, I asked my Google Home to play music for me. My Google Home decided to play the exact same creepy song that had been playing in the background during the creepy YouTube videos. This song was not at all similar to music that I normally listen to, nor was it similar to the songs that had played before it in the autogenerated playlist. I freaked out and I unplugged every device I own which contains a computer chip. These days, my main computer runs off solar panels and is not connected to the Internet. I compile everything from source code and the computer does not do anything unpredictable.
This experience was obviously terrifying. Had I been threatened by a criminal group? It seemed possible. Furthermore, I felt that the response of the police was strange. One possibility was simply that they did not believe me, and legitimately believed that due to my emotional state that I had a mental illness. However, another possibility was that I had indeed been threatened by a criminal group, and the police were part of that group. If I had been threatened by a criminal group, however, they did not indicate what they actually wanted from me. I decided that I would wait and see what happened and try to collect more evidence. If I had to call the police again, I wanted to make sure that my claims could not be easily ignored or denied.
2. The Paper
------------
If I had been threatened by a criminal group, it seemed to me that it was most likely related to my job, as that is the only significant thing about me that might be interesting to criminals. At the time of these events I was aware of at least one work event that was strange. In June 2018, my team at work published a paper on block floating point, an optimization technique for running AI algorithms on FPGAs (internet link, my photo). However, after this paper was published, it was decided to create another version of the same paper but with Intel branding (internet link, my photo). These papers are mostly identical, with the only notable difference being that the Intel version removes my name from the list of authors. I found this out after the paper was published online. The fact that my name was removed from this particular paper was especially frustrating to me because the block floating point technique described was something that had been my original idea. I thought of the idea and implemented it myself before showing it to the rest of the team. I originally called it "shared-exponent FP16" before I found out that there was a similar technique called "block floating-point" used in non-AI applications, and so I started using that name instead.
I do not know if the hacking attempt is related to the paper. However, both incidents left me feeling distrustful of others. If someone on my team was capable of removing my name from a paper in order to harm my reputation, what else might they be capable of? I have noticed that many people these days seem hostile towards me with no explanation. Why is that? Has my reputation been harmed in other ways that I am not aware of? It is a very difficult situation.
I was surprised to find that the version of the paper without my name was still online when I searched for it in order to write this journal entry. I thought it had been taken down after I pointed it out to others at work. I have been very reluctant to speak of this issue publicly, because I would prefer to avoid harming the reputation of others, especially given that many people named on the paper likely had no idea that my name was removed. However, the removal of my name from this paper was a significant event in my life. It is important to me that people are aware that this happened to me.
3. Corporations
---------------
In more recent years, I have had a lot of time to reflect on the nature of corporations. One thing that I have realized is that when you are working for a technology company, you actually sometimes have an incentive to complete tasks slowly, on purpose. Technically, whenever a person is making a salary and is assigned tasks, and they choose for themselves how long each task takes, they make more money for each task if the task takes longer. If a task could be done in a week, but you choose to spend two weeks to complete it, you get paid twice as much money for the same task. This incentive can also cause people to somewhat arbitrarily increase the complexity of their products, rather than work slowly. In many cases this is completely fine, especially when the work has an aspect of art or entertainment. You could argue that artistic choices are being made when designing user interfaces or programming languages for example.
However, I suspect there are people who explicitly intend to produce output more slowly. This seems like it would be more likely to occur when you are working on a product that will actually replace an existing product being sold by the company and you expect the same people who bought the previous product to now also buy the replacement. In that case, the company will make more money if they delay the release of the new product as long as possible, to allow people to buy as much of the previous product as possible. This is a similar concept to how a person receiving a salary might choose to purposefully complete their tasks slowly in order to make more money from each task.
In a free market, however, a company cannot admit to completing tasks slowly on purpose. If they did this, a competitor would emerge who was willing to do the same work more quickly and would steal their customers. For CEOs of technology companies, if they are intentionally working slowly, the rational strategy is to market their efforts as if they are working as quickly as possible even if they are not, and if a competitor appears, they acquire that competitor and convince them to match their rate of development. Essentially, the technology companies try to hire everyone who has a particular skill, so that the skill itself is monopolized by that corporation, and so the work can be done more slowly in a coordinated way. This is rational and as far as I can tell is the best way to maximize profit. At first glance, this sounds unethical, but in my opinion it is not that simple. Everyone in society coordinates in a similar way by not working on weekends, and I do not think that is unethical. However, at least in that case, the fact that people are working slowly is not a secret.
Now that I am thinking about things this way, a lot of things about how the economy works and how society works make more sense to me. In my opinion, whether working slowly on purpose is ethical or not depends a lot on what you are working on. If you are working on medical treatments or life-saving devices, it seems like you should work quickly. However, if you are working on something less important, and you are not discouraging others from improving the important things, then at least you are having a neutral impact and in a sense are collecting something similar to a basic income. This does not seem so bad. If we are actually on the verge of an AI singularity, it may be that all jobs will soon be automated and everyone will need a basic income. Maybe the corporations should hire everyone. That seems fair.
I am not sure if this theory about corporations explains anything with regards to my experience being hacked or my experience having my name removed from the paper above. I like to think my block floating point optimization was so good it scared the executives and they wanted to slow me down, but I actually have no idea. However, I feel that I was hacked, I contacted the police and was not believed, and so what recourse do I have left? I have to come up with my own theories.
Update (2025-07-11)
-------------------
After posting this information online, I re-examined the recent YouTube videos that I had printed on paper, to refresh my memory. After looking at the thumbnails, I realized that one thing I wrote above might not be accurate. If I remember correctly, the voice in the videos may have just been a recording of a person with a thick accent and was not necessarily AI generated. It is possible that the videos were normal YouTube videos that were uploaded for legitimate reasons. What was unusual was the order that the videos were being suggested by the YouTube algorithm while I was watching it. If the computer was being remotely controlled, this seemed like it could be a way someone could communicate with me while being plausibly deniable.
However, I have been thinking about the experience a bit more, and I have an alternate theory to explain what happened without it being the case that my computer was being remotely controlled. When it seemed like the YouTube suggested videos were communicating with me, I started intentionally always picking the first suggested video, to eliminate the possibility that my own choices in selecting the videos was part of what was causing it to occur. The algorithm kept suggesting videos with thumbnails that contained large words in text. The text seemed to sometimes be responding to things that I was saying out loud in my room.
I have recently been learning about Large Language Models (LLMs). LLMs are a type of AI model designed to predict text based on previous text in a way that can mimic human communication. When I selected suggested YouTube videos, I always picked the first video. However, there might have been other people like me who thought they were communicating with someone who was controlling their YouTube account, and those people may not have always been selecting the first video. The YouTube software probably uses some kind of AI model that is designed to predict the video that I might want to see next based on the behaviour of other people who have used YouTube previously. In a sense, maybe what has occurred is that other YouTube users have trained an LLM which is embedded inside the YouTube suggested video algorithm, and I was using the LLM in inference mode by always selecting the first video. I wonder if the YouTube suggested video AI model is complex enough to contain an LLM.
If I am correct, I think this is a very interesting concept, but it seems somewhat alarming. At one point it did seem like the algorithm was suggesting that I could lose all of my money. I wonder if that is a response embedded into the LLM because other YouTube users trained it that way. If so, maybe I am not the only person who had a scary experience.