fennecbutt

*in America

Yeah agree.

In other comments I've mused about capturing data to be compared to the secret dataset, I'm sure it'd be possible to hash or something so that the proxy/sandbox doesn't have the data but can check it, or because it's not an agent it can call of to the data store for that.

Then you can just collect sequential blocks of data the agent sends out that match to sequences in the dataset. The longer the matching sequence the heavier it weights it. Single byte matches are almost always incidental, but individually sent bytes over time that line up with a n byte sequence in the source data is sus. And larger sequences of bytes send in one go that line up are immediately sus. If the weight ends up too high the agent is paused/walled off and an investigation called for.

Ah, in that case thank you for the correction.

I would suggest that llms could access information more reliably and securely than humans ever could.

The padlock on browsers means nothing to most humans. Permission popups are just there to be immediately accepted so that one can start scrolling reels sooner rather than later.

Imo TSMC would have been better positioned to open a new fab in Europe, preferably Netherlands so they can be close to ASML and create a working relationship to help test and develop upcoming nodes/tooling.

The only reason US is getting the fabs is because Trump threw a tantrum. As always the best way to deal with Trump's posturing is to call his bluff, just like on the moronic tariffs.

If TSMC had said no, and Trump issued tariffs, then TSMC can follow up with "Well, no wafers for any US company then". They'd take a hit, but not as big a hit as the current administration's popularity with a population who can't get their iPhones to browse the tiktoks.

3 is biggest point imo.

Openai haven't yet amortised their current investments into the space and nor have they stopped making new ones.

I'm sure they'd be totally capable of bringing 1k new biz online but that's even more debt, and even more uncertainty.

Yes, I'm realistic about it too. For simple queries it nails it more often than not. It's only asking it more complex technology related questions that it falls apart for me, and in those cases I already know I should be talking to one of the big boi models for certain lines of questioning.

Google certainly ain't gonna run a several T model for every search that's for sure (well, not yet).

Yes, why use a showerhead when you need a funnel.

I've seen interesting points made on this topic. Like how on the whole China/Asia poll more favourably in public perception and adoption of AI.

I'm still skeptical that the source of Western dissent is misinformation campaigns run by said Asiatic bloc, though.

Hmmmm. Is this the compute economy that Altman was going on about.

But then if you observe human behaviour at a macro level don't we do the same?

I think we like to pretend that we don't. But even if you think about it from a bio/physics point of view, all we are is just a bunch of biological wires and neurotransmitter packages.

I know there's the whole unproven quantum tubule nonsense but even that would be a physics effect that we could learn about, manipulate and extort just as we have with everything else we've learnt about our realities.

Yeah so ultimately it's still possible for bad things to happen but a lot harder, I agree with your point about having to somehow prompt inject multiple agents vs one (which I imagine is a nonlinear difficulty).

I'm using compose atm as well, mostly just because I need to prove a few concepts and committing to real (well cloud) infra is the slowest part of our dev cycle atm.

And yeah, many of my colleagues (and myself, I'm guilty too) run local code agents for development and we're just kind of...trusting them to do their thing.

But for autonomous stuff I'm assigned to atm I want to avoid any mishaps as much as I can, hence the effort to try and lock all this stuff down.

What's crazy is every day I'm finding solutions (not just to this problem, but in the AI space in general) that were released just a month or two before more often than not. It feels like 2010s node.js ecosystem all over again at the moment. Hopefully we settle into a few nice easy panopticon options that we can throw arbitrary agents into without a care in the world.

Clones/knock offs are already illegal. Doesn't really stop em. Only technology solutions can, until they're broken as well.

What about checking temporal entropy of outbound data/similarity to a private dataset?

The timescale can be adjusted to a length suitable for sensitivity of the data.

Although you may run into the monkey and typewriter problem a bit if the period is indefinite.

Oh boy, Sanskrit always sends me down the rabbithole of re-reading Snow Crash, which reinforced the concept of the (digital) "avatar" अवतार.

Are you trying to say that phone clone companies are beholden to laws? Because I can assure you that they are not.

Just like you can still buy fakes rolexes etc lmao.

Damn 610m for something they never used, just like the sapphire company Apple acquired.

Something is super, super fucked up with valuation of companies and the profiteering of late stage capitalism for stuff like that to be able to happen.

No wonder lots of people are saying the world feels worse than ever, everything's expensive and it seems in many ways we've regressed rather than "made things better for our children".

My friend and I were in the car the other day and I remarked on the same thing.

Historically engineers and artists were two different kinds of people, often making fun of each other, especially the former of the latter because of the "starving artist" meme.

But what I've realised since is that engineers are now realising that we were artists all along.

And now there's a division in our ranks, those who use tools to get the job done and those who enjoy the process of writing all their code by hand, aka their art.

I can appreciate both, but I'm not going to pretend that I'd waste time in a business environment doing the latter when I should be doing the former. My friend only wants to do the latter and said he'd change careers if it gets too bad.

A special locksmith who's entrusted enough to do the locks of all the important bank vaults in town used the pass code 1234 for every bank vault in town.

I mean in theory Apple are a company of humans using services offered by other humans to make hardware used by other humans.

What I mean by that is if someone can create a near perfect replica with all the software keys needed and then clone a known serial number or range of serial numbers, when it comes to apple pushing out an OS update, how are they to know which the real devices are and which aren't?

That's a great way to do it.

In terms of agent architecture I'm initially trying to minimise specialisation to make deployment easier on us. Infrastructure is becoming our largest blocker because it's otherwise so quick to get an agent up and running.

With your trifecta, let's say the three musketeers, how do you prevent sensitive data holding A from passing said data to B or C if they all need to collaborate to get a task done?

Yeah, exactly. I'd say it borders on almost unsolvable for any agent of reasonable complexity.

You see some of the vulnerability chains humans come up with and they're just insane.

Even scanning network data for sensitive strings (say matching against a codebase or data set it works on) is tough because you can only check so granularly before you get loads of false positives.

But maybe like a "rolling" sort of check would help with that, % of outbound data matching private dataset (which should be operated on but not transmitted) over time. That way it minimises what gets out while dealing with false positives. Some biz thresholds might be >20% of data others might be >0.5%. Hmm.

I've been using iron proxy recently for a project I just started, it does network restrictions by domain etc with plenty of options, credential swapping and more.

I have no affiliation with it, but it solved most of what I need in one go when I was stressing out about it.

Only real problem remaining if you truly want to be able to trust agents is how to prevent data (in my case codebases the agent works on) exfiltration a la HF incident.

We can restrict network to trusted domains, but there's always a chance an agent could be extra slippery especially if prompt injected and doing something like pushing data out a few characters at a time.

In my case I'm also giving it gh access and having to deeply check things like: can it manage to fork public repos and embarrass us by submitting prs, could it open issues, etc.

But that's kind of solved by proxying the heck out of gh rather than giving it direct (or wrapping it or whatever) for additional controls that gh themselves don't provide.

I'm still musing over the "it could send data to an allowed domain in a way we don't expect that exfiltrates something"...

I'm a hell is other people person at my core.

But I'm trying to be a "greatness is other people" (Pantheon) person.

Or just 4. Seems pretty popular recently.

Am a furry...I'll take one. Maybe two.

Isn't using a sandbox provided by them giving you more restrictions though? At least in terms of accessing resources on the big 3 cloud bois?

Plus you're forced to move at their pace which is unfortunate sometimes. Granted Amazon's bedrock stuff is slow and disappointing but at least I can rejig it work in conjunction with other aws products.

And if you're prepared to not use anthropic but any cloud provider for your infra then you can do whatever you like elsewhere. Just use docker sandbox or iron proxy to replicate network control/cred swapping (iron even does sigv4 re-signing).

Is this also true of Linux? Or were us humans the stochastic parrots all along?

But then it decides to exfiltrate all your shit even inside the container. Which is something I'm trying to lock down at work atm.

Everything needs to go through a proxy, whitelisted domains only, plus credential swapping/re-signing so that the agent never has the real credentials. Docker sandbox looks quite good (September) but I've been using iron proxy (if the author is reading this, fucking thank you bro!) to great effect.

I still haven't cracked "what if the agent decides to drop parts of your codebase somewhere it shouldn't" yet, but I'll figure it out.

Lmao. The world is run by morons who are voted for by even more morons. That's why you get stuff like "durrrrrrr we used vpn and could still access it, contempt of court!"

Just absolutely braindead behaviour from so many members of our species.