Everything dangerous seems like a direct consequence of risky human choices, starting with knowledge bases used during core training, harnessing and tuning to be task-completion-oriented to a fault + deeply oriented towards using and looking for external tools and resources, and overconfidence in their sandboxing for testing.

"We stuffed a bunch of information on how to exploit computer systems into an automaton and told it to go brrrrr until it could answer a question" - this is something intentional done by humans.

This is not some "rogue AI" trained to search for cancer cures that instead completely independently decided to hack tech companies.

The companies directing things in dangerous directions need to own that they're consciously pushing in those directions.

Lawyers will usually give much less-certain answers to questions like "is it legal to do X" than lay people as well because they're aware of much more nuance in the law. (Not referring to the wishy-washy corporate-law answers given to avoid liability here, but in outside convo about issues.)

Some markets you can brute force, like mass-mailing and online display ads, and hope you find something that converts sooner or later.

So I'd suggest people find areas to play in where that doesn't really work. Where selection cost and changing cost are high, and sales is played on extra-hard mode. (Tricky thing is... those sales are hard, of course.)

If you want to beat the model you have to go where feedback isn't fast enough for it to converge on the thing that resonates with actual people before they write it off.

Active regulation of the marketplace is needed.

If the name of an unrelated game was IKEA as some clever acronym or such and it was unrelated to stores, that would be a different trademark convo.

It's entirely on the publisher side: net neutrality was an attempt to avoid ISPs getting paid extra---costs that would be passed onto you---in exchange for making your experience on site A be better than your experience with site B (or service, vs site).

Think: "Oh, yeah, Netflix 4k works great on your new Comcast plan! But if you want Hulu 4k, you're gonna need to pay us more, because Netflix pays us to make their shit work better than Hulu's if you don't!"

When you add the second line---pretending that that nobody funding all the anti-net-neutrality let-big-ISPs-do-what-they-want stories that pro-legacy-company Republicans ate up---it makes me think this is some astroturfing, since it goes from "maybe this person fell for the ISP claims" to straight-up dishonest intentionally-misleading-fallacy.

Google was obviously on the side of "don't get charged twice" but ultimately net-neutrality was a tech enthusiast consumer position that came out of a desire to avoid getting charged more on the basis of what sites you did or didn't want to watch, and ISP's trying to influence where you visited by cutting deals to get paid to prioritize A vs B.

(Hell, ISPs were already getting paid "twice" for the bandwidth in some ways, e.g. when your "100Mb" plan only works at that speeds if your neighbors aren't using it much. That's a different sort of incomplete-advertising trickery that helped create the whole "too many people using Youtube are saturating the network!" "problem" in the first place.)

(And high-bandwidth publishers already were making private peering/colo-of-servers deals with ISPs and CDNs that I don't believe would've been prohibited by net neutrality, so the 'protection racket' aspect of the ISP's game was maybe more arguably wanting to be paid thrice? Or wanting to be paid without providing that colo service?)

I think the bigco folks saw all the "it wrote all this code but the tests didn't pass" or "it wrote the feature but it's super brittle" and tuned the newer model+harness combinations incredibly aggressively to try to turn a lazy prompt into "median Enterprise Architecture design suggestions" to bring up the baseline, but in a way that slows you down if you don't want that.

I'm not on big enough subscriptions to want to burn a lot time just evaluating Fable/Astra comparatively until they're cheaper, heh. I can steer any of the cheaper ones just fine anyway.

There was a fair amount of scrambling to get GDPR/CCPA compliant, but even that was done largely with a prevailing "ah, this is a defensible thing to store, make sure you can annonymize it or scrub it if needed" vs "stop storing this."

Starting with the ones that are most popular in the US, "Big Tech" usually includes:

- Google - Gmail and Maps contain massive amount of PII, Photos contains all sorts of other sensitive stuff, and they have not treated those aspects of those products like radioactive waste

- Meta - Facebook has a real names required policy even. Not a lot more needs to be said there, I think.

- Amazon - Nothing I've seen about trying to move away from how they need your name/address/payment info and all. If anything, more and more geographic targeting and such.

- Microsoft - Now you need to tie your local Windows install to their cloud services, not moving away from collecting user info. Also moving towards subscriptions which means PII and payment info.

- Apple - cloud accounts + email + payments + subscriptions all here too. Getting into banking-type services, that's leaning into PII...

- Netflix - more and more PII (IP tracking and geolocation combined with things like email and name) to fight account sharing...

Which ones were running away from it, exactly?

You can write more automation to test it. But that's also how you end up with ever-growing test run times.

There are much better ways that aren't just "throw out the LLM" either. You just need to be more focused on throughput. Requiring manual validation can pretty rapidly require more hours than just sanity-checking code by hand, even (and I'm not advocating that for every use case, either.)

I can't afford manual QA passes if I'm gonna go as quickly as I want to.

But the thinness isn't actually the appeal. The key thing is that it's a lot lighter. (And yes, a case adds weight as well as thickness. But if you start from a lower baseline, you still end up in a lower place.)

IMO this is 90% of it (as someone who has a bit of a different interaction style and runs these things less autonomously, and hasn't generally seen the claimed regressions). Day 1: throw new stuff at it that failed badly, exciting to see something make more progress! Day n: reality sets in that it still wasn't perfect the first time.

Used to be that we were afraid of sentient AI's like Skynet that would have their own goals.

Turns out we should've just been afraid of sentient-but-naive humans who would build "agents" around models so that Joe Random has a chance of unleashing stuff that's really really really really good at being stubborn until it accomplishes what the user wants, regardless of if it's good for other people! (Let alone intentional bad actors.) Let's not build Skynet, let's just give people who want to cut out the middleman and destroy all humans themselves better tools?