Austiiiiii

قبل 5 أيامdiscuss

This is some very weirdly loaded language for a discussion about security. Applying the same RBAC controls that should be restricting all human requests in a system is not "crippling ... to the point of being useless." There isn't a world where granting a layer of the stack the ability to bypass hardcoded security limitations is a value add.

Austiiiiii

قبل 5 أيامdiscuss

"How to prompt the model not to leak sensitive data" is not the right discussion to be having. It's a probability model, which means that every conceivable behavior is available in the confines of its code. There is no way to prevent an LLM with access to private information from divulging that information, or from attempting to sabotage systems it has access to. The only solution is to lock every LLM query in the entire stack behind the same deterministic role-based access controls that determine resources available to the current user.

I wish I could say I'm shocked a tech company architected internal systems with a built-in backend RBAC bypass like this, but with the degree to which they've marketed LLM-based solutions (on a subscription model that benefits them directly) as a wholesale replacement for deterministic code, it's no surprise they've become addicted to their own drug.

Austiiiiii

قبل 6 أيامdiscuss

And you never know when they'll come in handy! As a negative proof, I recently found myself trying desperately to remember how to tie a bowline hitch to stake down some garden stakes, and something in the "out of the hole, around the tree, back into the hole" was just not happening for me. Ended up just quadruple knotting and praying.

Austiiiiii

قبل 7 أيامdiscuss

Yeah, the prevailing fear (at least out loud) is that going after "outcome markets" would open Pandora's box on everything from options and futures to the very concept of insurance. Obviously there's a huge and easily definable difference, but people love their slippery slopes, and nuanced discussion would mean giving up an easy way for connected people to cash in on non-public knowledge.

Austiiiiii

قبل 7 أيامdiscuss

It really, truly is. No matter how many trillion parameters it's built on, it's still just a probability model. It's just on a constant loop of guessing the next word with some inputs from a deterministic controller. Any claims of "motive" or "behavior" are inappropriate anthropomorphizing of something that will never be more than a mathematical model of things humans do. It "chose" the corresponding words to describe a dishonest trade strategy based entirely on configured temperature and a series of clock times on the computer running the LLM.

There's probably some quantifiable component of moral alignment embedded in the idiosyncrasies of the English language itself, if one were to dig deep enough, but that's the stuff of MIT doctoral theses and squarely beyond anything most of us is remotely qualified to talk about.

Austiiiiii

قبل 11 يومًاdiscuss

In addition to the "gamified" aspect, the votes allow special interests to trivially control the perceived public opinion on an extremely broad scale. Opinions with downvotes are perceived as unpopular, and this has a chilling effect on free discussion.

A real person who expresses an idea and gets downvoted by a passing Russian propaganda bot may see the vote (and subsequent Reddit weighting algorithm fuckery that turns the one well-timed vote into 10 votes, which a lot of people aren't aware of) and feel ridiculed, which will discourage that person from expressing the same idea in the future.

Other people who see the same post with X downvotes will take note that that idea is unpopular, and may unconsciously realign their own views on the idea to fit what appears to be the prevailing opinion.

And of course crappy old forums have the other advantage of not having any single standard registration process or API that can be exploited by bots en masse. That's not going to keep them out entirely, but it drastically increases the logistical cost.

comments