there's a lot in this comment and i don't want to engage with every little thing in there so lets focus in on what we are actually talking about: is AI theft?
the argument i've made multiple times is not that the technology as a concept can only be used to steal, but that the people developing, using and monetizing it are stealing.
I made sure to exclude individual users from this framing because the understanding that allows us to consider AI (not as a kind of matrix math, but as a technology primarily mobilized by huge incredibly wealthy companies with monetary interest in its deployment and use) as theft is unconcerned with individual users changing their relationship to creative technology and its output.
to restate that argument and get us back on track: AI (once again, not the matrix math, but the set of interconnected technologies built around a proprietary form of it that are being mobilized for profit by enormous incredibly powerful companies) is theft because the people in control of it steal. the closed technology of AI is predicated on open and free access to data for training without any pretense of compensation made to the creators of that data, the maintainers of resources like apis, servers and databases used to access that data or any expectation that the end model will ever be made public. the technology of AI is intended to be used to generate profit, and as such those in control of it have every motive to avoid compensating the people who created, hosted, cataloged and made available the training data or releasing the model to the public. it represents a leverage of public goods for private gain in very explicit terms and this understanding is even part of the pitch to investors.
Now lets look at the effects that AI (once again, the interlocking set of technologies used by google et al. not the concept of matrix mathematics) has had just in the past year because of this relationship to public data. hosts of historic information acting as public goods have incurred significant costs, sources of information once kept open have beed gated to prevent crawling and access to "content" previously viewed as simply fodder for ad draws has been restricted because of it's new status as training data.
That last one should have particular resonance given that reddit planned to monetize peoples posts and comments as training data to sweeten their IPO.
now we need to talk about the word theft because we both recognize that its' particular legal meaning isn't really applicable to this situation and that western legal verbiage has actually avoided enclosure in general. it would be very easy to simply take a pedantic reading of the word theft and then say "it's not theft for that reason!", but neither you nor I will do that because we are interested in discussing this new technology and the developments around it. in lieu of saying "monetizing the commons and taking actions that lead to their restriction" every time we need to talk about whats happening, we need a word that means roughly that.
historically, we can see many examples of the word "theft" used in that way, all the way back to the enclosures but even farther if we like. so because of it's historical use and broad meaning in the vernacular we can choose theft as the word to mean "monetizing the commons and taking actions that lead to their restriction".
Now that i've restated the argument and defined the terms, lets take a look at your claim that AI is no different than any other technology using your own examples:
The printing press did not monetize the commons or restrict them. an argument could be made that its widespread use lowered the barrier to entry for massive amounts of human knowledge and literacy in general, so it actually had the opposite effect. The automobile did act as an enabler for the monetization and enclosure of the commons through the development of the highway system and jaywalking laws, restricting roadways to those who could afford cars. so that could be a good example, but you don't invoke it, choosing instead to reference its effect on leather-working industries. word processors, either as physical machines or software programs do not monetize or enclose the commons on their own, however a person could make the (i think weak) argument that proprietary file formats are a form of enclosure. You don't though instead claiming that the word processor put secretaries out of work.
so, it seems like AI (hopefully for the last time: as a set of technologies leveraged to control and monetize public goods, not matrix mathematics) is significantly different than past technological advances but very similar to some events.