[eganist]

You're assuming power comes with an assumed guarantee of use. OpenAI has vast amounts of power with the data they're collecting, but the likelihood of OpenAI using it against any individual is small enough that an individual could consider it to be outside their threat model (I'm speaking using security language, but I doubt most people go so far as to threat model these interactions; it's mostly intuitive at this point).

Your family has limited power in the grand scheme of things, but the likelihood that they may leverage what power you give them over you is much higher.

The IRS has vast power and is likely to use it against you, hence why tax fraud is usually a bad idea.

Hence "planar" rather than linear.

[Zambyte]

> OpenAI has vast amounts of power with the data they're collecting, but the likelihood of OpenAI using it against any individual is small enough that an individual could consider it to be outside their threat model

I think your use of the word "individual" is a bit weird here. I absolutely find it likely that OpenAI is doing individualized manipulation against everyone who uses their systems. Maybe this would be more obvious if you replace OpenAI with something like Facebook or Youtube in your head.

Just because they are using their power on many individuals doesn't mean that they are not using their power against you too.

[eganist]

> I think your use of the word "individual" is a bit weird here. I absolutely find it likely that OpenAI is doing individualized manipulation against everyone who uses their systems. Maybe this would be more obvious if you replace OpenAI with something like Facebook or Youtube in your head.

> Just because they are using their power on many individuals doesn't mean that they are not using their power against you too.

Yeah but at this point you're identifying individual risks and grasping at straws to justify manipulating* everyone's threat model. You can use that as your own justification, but everyone manages their own personal tolerance for different categories of risks differently.

*Also, considering the published definition of manipulation is "to control or play upon by artful, unfair, or insidious means especially to one's own advantage," I think saying that "OpenAI is doing individualized manipulation against everyone who uses their systems" is an overreach that requires strong evidence. It's one thing if companies use dark UX patterns to encourage product spend, but I don't believe (from what I know) that OpenAI is at a point where they can intake the necessary data both from past prompt history and from other sites to do the personalized, individualized manipulation across future prompts and responses that you're suggesting they're likely doing.

Considering your latest comment, I'm not sure this discussion is receiving the good faith it deserves anymore. We can part ways, it's fine.