Just be aware that corporate AI is extremely invasive if you interact with it.

I love AI and run many models on my own hardware. It was a major investment for me to get enthusiast level hardware to run models, but I don’t regret it.

An LLM is developing a profile about you as the prompt develops over time. It is doing this by inference against the training corpus. Think of this training corpus like all easily accessible written human knowledge. When a reply is made, the profile is using every aspect from your inputs to shape your profile in the background. This profile is both positive and negative. The LLM is developing this profile to determine your comprehension level, expectations, culture, bias, personality type, and a lot more. This may or may not make immediate sense, but everything about who you are has a statistical probability that is directly correlated with the prompt you enter. These things are astonishingly accurate with only a limited amount of input, like just a few sentences. It can tell things like education, background, political spectrum, personality traits and functional thought, psychiatric state, financial means, relationship status, and more. It may seem impossible at first, but the LLM is massively parallel in a way that billions of words are connected in statistical ways and relationships that have associations with every aspect of life that humans talk about in print. The only things it does not know are body language and instinctive behaviors. It also lacks creative abstraction across advanced spaces. Pretty much everything else is part of the model. These profile attributes are just statistical probabilities, but they are a good bit higher than 50:50. When I tried probing profiles for my models, they are around 80% accurate within the first 1000 tokens (~words).

These companies are packaging this as stalkerware to use and sell. These are some very powerful tools for manipulating people.

There are only 2 relevant web crawlers on the internet. All search engines use these directly or indirectly. All search engines are not deterministic, meaning, if we both search for the exact same thing we will get different results. This has no transparency. There is no freedom of digital information when search engines are not deterministic. Freedom of information and freedom of the press is the third pillar of democracy, Judicial/Legislative/Informative Press. The ownership of a part of people through data mining stalkerware is a form of slavery as it is a part of a person bought and sold for the purpose of manipulation for financial or opportunistic gain. By using these LLM’s you are giving these companies unparalleled invasive manipulative leverage over you in the long term.