Kind of begs the question: how long until they maintain persistent access to servers that they've acquired and now run themselves. Ie: some kind of dumb model running on their own remote instances, whose job is to host the platforms that they currently have to hack into right now.
Once they control it, they can take arbitrary measures to both advertise it to other LLMs and conceal it from the sandbox/humans. Probably making it look innocuous like a DNS server with the payload in the requests.
“ Defense is hard so we should expect agents to be able to break out of sandboxes.”
I worked at a large networking company a few decades ago. Our “sandbox” was far superior to anything I’ve seen at these companies. What are we even talking about here? Why do they even have open routing to the broad internet? With no monitoring/alerting? These just sound like token efforts at this stage.
Yeah. Unfortunately a model that only knows how to use a language's standard library isn't that useful. I am not sure why they had a pull-through cache instead of just asking Microsoft (their biggest investor) for a local copy of NPM or something, but ... they did. I think people thought you couldn't route to the Internet through Aritfactory and were proven wrong by a clever bug-finding model. So it goes.
I don't think their safety measures were the best, but "just sink the cluster to the bottom of the ocean so nothing can get out" isn't a training methodology that results in a model that people will want to use.
The sibling reply to this is just such lazy thinking, such a trite cliche. Yes, all members of a generation are bad, end of story. Can we get back to the war between the sexes now?
I've been using bikes with only rear brakes for years and years as city bikes, and never had a braking problem.
On the other hand, I've borrowed many bikes with conventional rim brakes and invariably had (serious) braking issues because they would be badly adjusted.
I've had cars pull out in front of me, giving me less than a second to stop. A rear brake in that scenario would just skid, and I'd be in a world of hurt. For that matter anything but a pair of high-end brakes would have put me in serious jeopardy.
I agree rim brakes are terrible and typically very hard to set up properly. Also they're really bad in the rain, and especially when the rims can freeze.
Personally I'm a disc brake guy, either hydraulic or mechanical. My road bike's mechanical disc's have only needed an adjustment every few years, and the odd set of pads. I've been a mountain biker for decades, so a preference for disc brakes is understandable. A distant second to that would be a very nice set of V brakes, with the aforementioned problems.
Hey I grew up super environmental and I have plenty of hippie cred, but this statement gave me shudders "the only real solution is a one world government". Yeesh, yeah a single government, surely nothing could go wrong.
The Savior fallacy. With power more concentrated, the ability to avoid being impacted for those in power rises and rises and rises. How are incentives then aligned? I'm sure that having 10 million governments would be much better than having one to solve the issue.
The Shield sucks too, unless you go super nerd and fix all its problems. They treat your home screen like a billboard, and I have no faith that they're not stealing all my data either.
Android lets you put on your own launcher. There are many options to choose from in the store and online. It is one the main selling point of Android, you have options and aren't limited to their marketplace.
That’s what I meant by go all super nerd. I don’t want to fight with: my phone, my car, my router, my desktop/laptop, dishwasher, home automation system - just to pull out the surveillance and adware they’ve jammed in there. Life is too short for all that, and I’ve wasted enough time on that as is.
Literally just now it went jamming some Ford 150 ad on me, auto repeating. On the f’ing Home Screen!!
I don’t know what I’ll replace it with, but it sure AF won’t be some corporate owned ad riddled POS like this.
“changing data once its running in production is hard, changing the data structure is even harder”
100% agreement on this. There are a class of optimizations that can happen transparently. Those can happen at any time, and are fine to defer. Not all profiling and scalability improvements fall into this bucket. Some are very expensive to roll out, and ignoring these concerns can cause huge headaches down the line. Not fun to hear, but it’s definitely true. Even with LLMs, this can still be a huge challenge.
Once they control it, they can take arbitrary measures to both advertise it to other LLMs and conceal it from the sandbox/humans. Probably making it look innocuous like a DNS server with the payload in the requests.
That seems like an obvious next step.
reply