The data I see from iKnow is a sloppy clusterfuck, I suspect from botched syncs and lazy bookkeeping on my part, and the fluctuating quality of Apple's "services" over the years. My Mac settings probably go back to the 1990s and Apple periodically did a bad job. Pain points I remember: Google contacts being separated and then with a new phone not syncing correctly, Mac address book burping on contact avatar images, importing lists of contacts from short-lived 3rd party apps, shared Photos fucking shit up, the iPhotos upgrade fiascos, transient contact info for my kids, moving onto/off of Gmail, etc.

If the FBI got ahold of my (and others'?) iKnow data, a lot of people might end up in a sloppy jail with a sloppy judge sloppily applying a sloppy law...

> Keep activity

> Keeping your activity lets you pick up chats where you left off anytime and helps improve Google services, including AI models. When this setting is off, Google still saves chats for 72 hours to respond to you and help keep Gemini safe.

Why is there not yet fully open models made by/for the FOSS community (as opposed to open weight by entrepreneur-style companies)? Cost of training? But what is this cost really?

- Do not include costs of running model in cloud - Do not include IP costs - Do not include salaries, marketing, etc. (all these are separate expenses, different depending on individual goals)

There would be pre- and post- training "systems" that could be repeated.

Training data/acquisition could be separated out (so IP-liabilities could be taken on by individuals).

The idea would be a system for small groups/individuals to snap together the different aspects of training (data, pre-training, training compute, post-training) and to train their own (and share).

Won't this soon come down to a few thousand dollars for models that run on modern consumer hardware?

But I've slowly come to realize that AI+"dark resources"(scare quotes)=something bad! And this is imminent (maybe I'm naive and it is already ubiquitous).

We see here, with Kimi, that there's growing traction with the idea of eliminating types of training to avoid evil. I'm skeptical of that.

And then we'd also do that for all kinds of potential terrorism, sex-crimes, political subversion, and evil advocacy?

It would be the "restricted section" like in the Hogwarts library?

Deliberately keeping all models ignorant doesn't make sense. It won't work (whoever breaks the pact has a massive advantage); it won't be enforceable; it will have downstream impacts on things you don't want it to be ignorant about; this kind of regulatory regime invites adding more and more to the "no-no list".

Is it not possible to focus on hardening? And on new legal liabilities for operators who do security attacks?

So there are these "resources" out there: massive botnets, stolen credit cards and credentials, etc. These often seem to be linked to or supported by state actors.

How can we hope that these "standing reserves" aren't used generally by AIs as if they are just another "skill" they know how to leverage?

Our infrastructures need serious hardening.

The Saussure idea of the differential nature of language seems relevant here too: words aren't just lists of things you look up, they are connected to each other and distinguish themselves by being slightly different from the other elements of the clusters they belong to.