Real-life experiences behind the problem this project targets.

Point 1

Text datasets for emotion / behaviour analysis are poorly labelled, for example:

The main problem is that the labels are often wrong and biased towards particular classes. Maybe generative models cannot completely solve this, but run many of them concurrently with Bayesian statistical methods, and they just might.

Point 2

Working in data teams, every time I get to see a fun project, I have to do the boring tasks first, for example:

  • Data scraping (XML)
  • Data Analytics
  • Debugging ETL pipelines

Sometimes there are surprises, like the workflow crashing on a Monday morning. These system failures may never stop entirely, but they can be reduced by a bot that mimics me exactly.

When a data pipeline is built incorrectly, it can lead to many costly mistakes, but that doesn't mean the programmer is at fault, and no one can know for certain. One can imagine that if a programmer were as devoted to the problem as he is to his wife at home, surely the problem would stop surprising him with another man (open ports).

The current solution entails running multiple daemon processes with unstoppable life cycles; without them, Apple devices wouldn't be here today. If daemon processes could think, adapt, learn and make decisions on their own, then maybe we would have more time to think about whether we are truly alive.

I once asked the world why this buffoonery of tools had not been automated, and everyone assumed I was just a Padawan who did not know any better. Now that I am older and a little bit taller, I am going to ask the question again: “How has this not been automated?”

One way to gain market exposure is by leveraging social platforms like Reddit, Facebook, Instagram, etc. Out of boredom, I've attempted to manipulate these algorithms through botnets, i.e. temporary test users. I am well aware this is not an ethical solution, but I do it anyway, and I am proud of it. That being said, I am proud to be part of the rare percentile of users who have been flagged, banned or revoked from platforms, including League of Legends.

I see these failures as lessons, and I use those lessons as feedback on my programming skills. Sometimes the consequences worsen, but on good days, my botnets live a little longer. I am an amateur, so I make errors like missing control variables, such as the peak hours of active users. Should I analyse the causal effect, or apply correlation-based inference to this feature? The answer is simple: just do it, i.e. pick one -> negate one -> repeat until you are left with one option.