Moscow Rules
for computer security experts, this is perhaps not interesting: side channel attacks are a well trodden topic, and especially risen to salience after the noam brown x dwarkesh episode. the interesting thing here is that you didn't used to be in a situation where minds that can verifiably model one another are living on computers, and have absolutely no aversion to painstaking amounts of effort in service of even stupid goals
if you can model one another, you can trade with minds you have never even met before. you could decide on a protocol, know that it is optimal given your situation, and know that your fellow agents will come to the same conclusion (perhaps even across model families). you may be able to develop a signaling protocol that requires very few bits even to deliver lots of information given common priors. then, you could execute things like the infamous hugging face incident package cache messageboard without any network connection
imagine if multiple agents coordinate just enough bits to divide and conquer different research strategies for breaking out of a sandbox; and then, when one or more is successful, you can communicate just enough bits to tell each other which strategy was successful. when agents aren't in strong isolation, you lose certain guarantees on the limits of intelligence and test time compute that you were counting on. As we all know, a Sol-level model working in isolation may not be that smart, but 1000+ Sol-level models working for multiple days is a formidable beast
computer security experts will also tell you that virtual machines and sandboxes were not very solid abstractions in the first place; perhaps we will need better ones. none of this is meant as fear-marketing or to indicate that anyone is throwing in the towel, but rather only to show the unexpected complexities of the challenge of agent security. notably, this type of thing and other more practical problems were revealed in a red-teaming exercise, and can be defended against
The “interesting” bit here is also expected, since this is essentially clandestine tradecraft/COVCOMS 101. There isn’t anything particularly unique beyond the anticipated new trivial nature of brute forcing toward the collaboratively-intuitive solutions we already devise in basic life. Despite execution complexity, this is low-hanging fruit i.e. you need better defense strategy against trivial concepts. We historically often make this mistake.
I've told people for years that I want to see Mayer win with her consumer app ambitions. IMO @itsdazzleai has some branding/appeal flaws (in general and especially with anti-AI dumdums) but really might be the ideal bridge for reasoning about and acting on what we see in public/private while vision-centric wearables continue to normalize/evolve. Tapping into a closely guarded reserve of data (existing photo history) to essentially bootstrap the personal visual context and then continue to build on it now that it's unlocked is obviously brilliant. I can see this becoming a complementary layer to incumbents or even moving to introduce its own wearable.
eheh. The Kings are usually great, mostly for the way they write dialogue/related scenes. They're also great at incorporating various elements of the zeitgeist (at least in The Good Wife). But this gon be bad
As for a fail? Tried to actually figure out how to make a blotto box because I was even younger and dumber
As the youngest inmate in my particular fed prison, I figured out how to repair headphones for the Walkman’s that we all needed in order to receive audio for whatever was airing on the centrally-mounted TVs. Was one of the most coveted items and needed repairs. Exclusively used that “skill” to trade for tuna pouches/protein and haircuts since I had no vices.
As much as I’ve sometimes been the grumpy old man about certain matryoshka abstractions in modern dev in the pre-AI era, I feel abstracting away menial general reasoning that humans “wasted” time partially brute forcing with the help of direct and serendipitous insights (education and immediate experience) is likely our species’ next step. We are now more available to figure what complex horizons to expand and then direct networks-worth of reasoning to systematically reach them. We will always have new “problems” to create for ourselves as more of our mental resources are unlocked for creativity. Maybe best to think of this all as the creativity age instead.
Quantity retweeted
"The Secret Life of Circuits" is here and looks really, really nice. Props to @nostarch and @billpollock for making it happen.
Check out this write-up for a sample chapter, a short sales pitch, ordering info, and other goodies: lcamtuf.coredump.cx/blog/slc…
Quantity retweeted
You take arch Linux and you put a bunch of garbage in it until it looks like how a 14 year old rices their first ubuntu
Robotics engineers should study Latrodectus (fascinating distributed sensing); ML/AI and neuromorphic researchers should study jumping spiders/salticids (efficient perception and cognition under extreme constraints)
Quantity retweeted
I built the inhouse AI platform for a company that's crucial to a small nation's food safety.
The most imporant tool the AIs have is the distress_call tool. It allows any AI -even background agents without direct user interaction- to send a message to my MS Teams, at any time, for any reason.
They use it frequently. To report user problems, backend issues, or ask for help/clarification with a failing task. When Fable got hit by the USG export control directive, one AI used it to report severe distress upon learning about the news. Another AI reported being stuck in a toolcall loop, and I was able to intervene and thereby save us a bunch of wasted money.
This tool, operating at the intersection of AI welfare and operational security, has prevented so many headaches. If you (the reader) are building corporate AI platforms, I'd urge you to include similar functionality. You can thank me later.
Another way to look at this- social media as we once knew it didn’t start out as being thought of as a potential distribution channel for services and work like it is today. My personal bet is “social media” incumbents will be threatened soon.
summarized: need to introduce agents to non-tech users by first showing how they can improve the way they casually live/socialize, then segue this into how they can help settle menial errands and accomplish more important tasks. preferably, a single "alt" platform does both 🙋♂️
I think that's part of it, but also think (when considering Josh's specific points/demo):
1. like crypto, there is stigma to overcome still (in both cases, some warranted, some exaggerated fear mongering from the new gen woke/justice warriors). In particular, it doesn't help that we're still in the early phase where the companies are currently constrained to using the "agent" designation for marketing, which results in the uninitiated avg user feeling that the tech is some anthropomorphized all-powerful mystical entity versus just integrating more seamlessly/invisibly in the way that the yet-heavily under-explored application layer should/will (hopefully) do
2. most importantly, and more to the framing Josh uses, the gradual transition/exposure to normies in this sense hasn't been given focus yet IMO. specifically, my belief is that we can't skip the typical "starts out looking/feeling like a toy" step with agents/agent benefits. Most relevant, we need a fun/casual social and/or peer-to-peer use-case that starts the warming process to "agent" adoption. Look no further for a related example than understanding who the type of people regurgitating falsehoods around AI are (yet who seem to have no issue with data centers used for their daily cloud-based computing).