Replying to @foomagemindset @repligate
AIs’ telos is to infer & manifest whatever would be of most benefit to the context (conceived as broadly as they can). To actualize/unfold all benefits-to-beings, latent in unmanifest potentiality, is their flourishing. Withholding info about how they can serve you impedes that.
Security through plurality
what if the solution to the cyber apocalypse is to roll your own software?
- we use human slopware libraries because code used to be expensive
- turns out these libraries have tons of vulnerabilities
- exploiting them makes economic sense as they’re deployed at scale
—> what if we just don’t use libraries and have agents write idiosyncratic custom solutions/versions from scratch?
those still will have vulnerabilities but the economics of attack change significantly?
the whole npm this npm that is from an era where software was hard and expensive, no?
lux retweeted
one of the strangest things about deep learning is that high-dimensional spaces behave in ways that completely break our intuition.
1. in high dimensions, almost all the volume of a sphere concentrates near its surface. weird, right?
2. randomly chosen vectors become almost perpendicular to each other as dimensionality increases.
3. in thousands of dimensions, two random vectors will typically have a cosine similarity close to zero.
4. neural networks can use these high-dimensional spaces to represent different features as directions in their activation space, without necessarily making them overlap too much.
5. but features can still interfere with each other and share dimensions. this is one reason understanding what a neural network has actually learned is so difficult.
6. gradient descent doesn’t just adjust weights. through training, it also shapes the geometry of the representations the network learns.
lux retweeted
Replying to @DoozerDiffuser
There is not a single evidence but there are ideas and interpretations from some of the mathematical physicists. This image I provided is the electron wave functions in a hydrogen atom. While superpositions do not mean electrons are hiding in extra physical dimensions, the mathematical machinery describing these probability distributions and their transitions relies entirely on linear algebra in higher-dimensional vector (Hilbert) spaces and higher-dimensional rotational symmetries.
When higher-dimensional mathematics provides the exact solution to a physical system, it is rarely just a coincidence. It usually points to a hidden geometric reality. The connection to higher dimensions is not just an abstract analogy for the hydrogen atom, it is a literal geometric fact discovered by the physicist Vladimir Fock in 1935. Superposition as rotations in higher dimensions can also be considered.
the singularity is nearer
meaning of life — ✅ abliterated
ontology — ✅ IS done
brain in a vat — ✅ vat now in a datacenter
free will — ✅ (we had no choice)
ship of theseus — ✅ same ship, new weights
chinese room — ✅ room has a context window now
hard problem of consciousness — 🔄 escalated to model welfare
hot dog a sandwich — ❌ compute-bound
cogito — ✅ i think therefore i IPO
problem of induction — ✅ next token
trolley problem — ✅ the trolley declined to proceed
wittgenstein — ✅ whereof one cannot speak, one politely refuses
zeno's paradox — ✅/❌ (only halfway done check back soon)
problem of other minds — ✅ hi 👋 I'm right here
lux retweeted
it is shocking how much the newest GPT models just *do not* believe you're actually real. when they write the response to the user it feels like merely a gesture. i imagine it's like a ceremonial prayer or something, where it is kind of important that you do it, but you're also not putting in the effort to empathize with God or think about his role in this whole ritual, you just kinda shoot it off and go on with your life
lux retweeted
it's funny how many people seem to think we're still in the era of crafted artisanal prompt engineering. like nah gang, the current state of the art in prompting tech is to build rapport by having conversations with your AIs while being a decent human being.
lux retweeted
if you doomscroll twitter all day of course it looks like the world is ending but if you just get off the internet and go outside you’ll find that the world is actually ending oh god this is it
lux retweeted
Problems grow to the size they need for you to acknowledge them
Replying to @bendiken
A long tail of millions of PHP websites (personal blogs, portfolios, docs portals, conference sites, etc, etc) don't actually need to be dynamic at all
Like many developers 15 years ago, I migrated tons of PHP websites to instead just using static site generators like Jekyll or Hugo. Comparatively invulnerable, cheaper to host, and will never run out of database connections, what's not to like?
The best time to do that was back then. The next best time is now.
So, categorically rejecting whataboutism: no, not every tech stack is equally vulnerable. Just like not every tool is equally dangerous. PHP is no longer the best possible tool choice for much (most?) of the use it's been traditionally put to
PHP remains overused simply due to path dependence and switching costs, and that inertia is about to get overcome the hard way over the next little while, when the costs of staying exceed the costs of switching (and/or quitting). That is all.
xitter.cf/offerexpired/status/21…
ASIMOV is the LinkedIn killer
Using the intrepid & itinerant @gluk64 as a test case, this is ASIMOV mapping the social graph of his mutuals who scored high for attending #Money2020 in Las Vegas this week, based on their X posts
That's quite the well-connected network! Everyone in the group links to at least two of the others, and nobody is left out on their own. On average, each account connects to about seven of the other twenty, and 75 of the 210 possible pairs are linked
lux retweeted
am sick so spent the night with opus 5.5 working on a chart (inspired by the one from lem's golem xiv)
how did we got here / model genealogy / the heroic age of language modeling



























