pona-a

I'm currently learning electronics, as well as having just recently sat my Physics/Math finals. I can tell you with 100% certainty you CAN'T learn shit from /just/ video.

Actual learning requires you to think about what you just read, maybe re-read it multiple times, stop to try and solve a example problem, etc - all of which require you to stop/rewind which video inherently disfavors.

Besides, I think just like handwritten notes might have a slightly different neurological effect than typed ones, reading might just be a very different mental muscle more connected to comprehension; humans had oral language for much longer than any script, so maybe it came with some different connections to higher brain structures as well.

That a government has a much easier time controlling who's renting a building with the energy consumption of small city than what everyone's coding on their laptop?

RATs do not require a datacenter to develop.

Because there is no more tech industry. Only gambling.

And Qt. They even made QML with literal reactive JavaScript.

There are a few other banks running their own NFC payment systems, like Swedbank in my country.

I recall a year or so ago, there's been a story about someone hacking McDonalds loyalty program, with that app doing something stupid like storing your balance on the client or something. It seems instead of firing whatever offshore sweatshop that made that, they just doubled down on "mitigations".

Play Integrity has several levels. GrapheneOS MEETS_BASIC_INTEGRITY, which I believe only requires a locked bootloader and no superuser.

There's also been some discussion of spoofing MEETS_DEVICE_INTEGRITY, since before Android 13 it didn't rely on a TPM, and many apps don't want to lock out older devices, but it's been decided against it [0].

[0] https://github.com/GrapheneOS/os-issue-tracker/issues/1986

I kept thinking this was a play on crank as in charlatan and crank as in hand-crank. Generate any flavor of crankery with the turn of a crank. Turns out it wasn't satire...

I do not defend the current state of things where a select few companies get to shamelessly violate the law with the entire legal framework bending around the weight of the money trapped in this speculative bubble.

I believe LLMs are at the very least an under-researched technology or less charitably, an ongoing effort to strip intellectual workers of their rights and privileges.

What I am saying is the reasonable demand for attribution runs counter to the nature of these systems as we know them. There is no magical "release the attribution" button Anthropic could press if they wanted to. Unlike per-state taxes, are actual PhDs working on, at universities and private labs, because transparency has been the public number one demand since day one, and yet all that exists after 4 years of funding are only the first incomplete steps.

The most likely outcome of imposing this obligation is commercial LLM providers quickly folding, finding a loophole/displaying false attribution, or settling for notably worse performance. That is of course not counting how these companies will be on the hook for a civilizational amount of licensing fees.

(Per the DRM point, I believe we can agree the goal of simultaneously displaying a piece of media in the physical world and somehow protecting the viewer from storing it is effectively impossible, without hiring a trusted guard to hold the viewer at gunpoint if they dare touch the trusted viewing apparatus or pull out their phone, at least in its strict form)

I am personally okay with shutting down an industry that cannot legally exist in its current form, especially one so openly hostile to every field of human endeavor. But no matter your position on that, we must keep in mind no "ethical" or "legal" AI industry can exist without making either adjective meaningless.

Is it actually possible to determine how much the weights were influenced by each work?

I might recall reading some interpretability paper years ago that trained a special model that could attribute each answer to a part of the corpus (like Wikipedia, ArXiV, or "Blogs") but it had a non-zero effect on performance and wasn't nearly as straightforward as weights go in, attribution comes out.

The funny thing is most of the world had already pioneered the airhorn ID long ago. Very few of them saw any decrease in 4 AM airhorn activity, yet some are already well-known to arrest and harass airhorn users to international human rights observers' condemnation.

We are just animals, more capable in reasoning than many others. LLMs aren't; their subjective "experience" restarts every session, they degrade into incoherent slush with larger context windows, and their reasoning is real but extremely weak, only compensated by the a civilizational corpus of information compressed within. I concede we scaled them enough to discuss them as tools that can produce useful outputs, but claim this echo of humanity is itself conscious by any non-trivial definition is either a product of psychosis or some long-standing hatred for life.

Congratulations, you have turned CUPS into a long-term support contract with Anthropic at $20/month, except the other party doesn't have to actually fix your shit and can arbitrarily alter the agreement.

A modern edition desperately needs an AI chapter

Only by having the LLM random walk the hypothesis space with a validator rejecting invalid ones.

The reason why LLM hypotheses are any good is because it already consumed a civilization worth of knowledge. You couldn't have bootstrapped such system with nothing but a few priors/axioms and let it discover the universe.

There's also the matter of brand reputation. You don't want to make your ads seem like your own or your user's public communications.

Homomorphic encryption doesn't exit at any meaningful scale. The weights must still be stored plain in the RAM for inference to work

Oh don't you worry your little head, it's coming in the next patch

It's simple to find out. Add Claude /etc/hosts and see how long you persist.

Attention is an O(n^2) algorithm. Combined with Moore's doubling, it will at best produce linear growth (assuming Moore's law is still remotely close to alive)

Why is U-Block telling me stonetools.ghost.io is in the fishing/badware list... If the webmaster is here, you might want to take a look.

A vibe-coded EFF Cover Your Tracks. The fact this made it to front-page is spookier than its contents

Believe me or not, that's good news for me. I actually really enjoyed your writing, and I'm glad that feeling wasn't misplaced. I'm sorry if my remarks came across as mean-spirited...

Pangram agrees with you. About 25% of the text trips the detection threshold, mostly towards the later half.

I don't want to make any accusations, just give some evidence to the above comment.

26% of the sentences tripped the detection threshold for their classifier. That isn't a detection probability on the entire text.

26% isn't high and it's concentrated towards the end, so it could be the author just used it to revise some later paragraphs.

This essay returns 50% on Pangram, which matches my impression of it. The human-written bits do have real insight, but it's being buried in vacuous slop that makes it.

I understand from personal experience how addictive writing with AI can be. Universally when I looked back at the result a week after composing it, I saw it was what it was - slop - and did the responsible thing - throw that out and use my own brain to express those thoughts, or decide that thought needed more time in the oven before touching the page.

If the original author is reading this, I beg you to consider this instead of becoming another contributor to the internet's cacophony of slop.

Projects like this are generational achievements, and we should demand public funding be given to them. It is a great loss its writer had to step down for 5 years, being too busy with an industry job, when they could have been so meaningfully advancing CS education.

Loosing their #2 market instead of absorbing a modest compliance cost?