quotemstr

As an aside, it's just silly to make the repository a bunch of checked-in patches. Git already does revision control. You don't need to do revision control in your revision control.

Fair cop.

Would the author feel the same if git had used MD5 instead of SHA-1?

Big model smell is a real thing. For certain classes of problem, ones you get a feel for but can't easily articulate, a big last-gen model can get you what you're looking for when no quantity of tokens from some ultra-RLed mid-size latest generation model can.

Outlawed? Only in safetyist dreams

These providers can also just ignore insinuations they're being "unethical" when people come to them with steering vectors in hand. Nobody has to listen to the scolds.

Because P!=NP (very probably IMHO) there's a huge class of problem for which LLMs are useful on the expensive and heuristic-y generation side because the verification is relatively inexpensive.

They brought it on themselves by marketing a read-for-free product

They're not doing autoregression, so all the outputs are computed in one big forward pass. Very cheap.

It looks like a specialized encoder-only(-ish) transformer with scalar and ordinal output heads. Acausal in effect, maybe? Probably not even autoregressive?

I'd use this as a tool an LLM can use for specialized tasks. It's not AI in itself.

"Stronger" is absolute bullshit. No data race prevention. No intra-object protection. Mandatory GC, and 5x the compute cost. Nothing like a statically safe language.

Shame that you can't then link /proc/self/fd/N to a different node in the filesystem namespace. That'd be neat and symmetric.

What are you going on about? Why shouldn't there be some temporary authorization prompt for clipboard access? You could even wire it up to control-V somehow such that the same keystroke would grant temporary access and communicate a paste to the application. Nobody is talking about making the clipboard unusable.

No.

The wheel of fashion turns once more.

I don't think GitHub or any other private party gets to decide what attaches to the perjury clause. A contract can impose liabilities, but it can't create a crime out of thin air.

The perjury part of the DMCA doesn't do what you (and most people) think. It attaches only to the DMCA claimant's authority to act on behalf of the copyright he's alleging is being infringed. It does not attach to the veracity of the allegation itself.

For example, the copyright holder for Bad Apple could face a perjury charge if he issued a DMCA takedown request alleging that something on the internet violates the copyright to Snow White (which he presumably does not own), but would not risk a perjury (in particular) penalty for issuing a takedown request for "Stairway to Heaven", the King James Bible, and the Linux kernel alleging that these things violate Bad Apple's copyright.

You don't have to quantize all layers and all dimensions uniformly, FWIW

This is going to get about 0% uptake. People's preferences, as revealed by their behavior, align with the algorithmic feeds. That's the whole point of an algorithmic feed.

I don't know what all that's about, but the day-to-day experience is that you can find a wide range of opinions on X and a much narrower range on alternatives. I don't think a lot of people who aren't already on these alternatives are in favor of restricting the range of expressed opinions.

You're not going to get many volunteers to move from a free-speech platform to a censored one. Everyone who wants the censorship is already on BlueSky or Mastodon. Part of sharing a society with people is being okay with their being able to say things you don't like.

We could never rely on scratchpad to reflect true model deliberatons.

Keep in mind that LessWrong a community that 1) ostensibly wants to slow down AI so we can prepare for it, and 2) in actuality, wants to make the Bay Area EA community a kind of guild that controls everyone's use of AI.

Take anything they write with a big grain of salt. EA writings these are mere apologies. The conclusion is preordained. Authors start with the goal of slowing AI and work backwards from there, trying to see which arguments resonate with the pubic. You can't unsee it.

Not everyone in the AI space approves of these people or their doomerish.

It is the real thing. Not "almost". The actual thing. If you want to forestall accidental heap policy, make a storage policy that prohibits it.

Ah. Dyno (https://github.com/ldionne/dyno) is good at this stuff. If you have types Base, Child1, Child2, you can just return a Dyno object that can be any of these, expressed as a tagged union and not a Box-equivalent, and then do regular vtable-based or otherwise polymorphic dispatch into the object. You can also arrange it so that if you have a Child3 that can't fit in the (Base, Child1, Child2) union, the Child3 can be heap-allocated and invoked transparently as well. It's open-world type erasure.

C++ is so freakishly powerful is that it can not only solve this problem, but it can solve it via a regular library and not a language extension.

"We"? Yeah, okay, Your Highness.

As for the content: the law says what it says. Where in the constitution of the United States of America and its amendments does it say that a state government must be truthful about public health statements?

The law says what it says. A behavior might be rotten, evil, illegal under other documents, and an all-round bad idea and still not a violation of, in particular, the constitution of the United States of America and its amendments

Where is the constitution does it say the state can't do that?

What does your explanation have to do with the fact that C++ can express by-value returns of complex objects?

My question stands.

Has anyone actually been charged and convicted for disclosing knowledge of a vulnerability in exchange for money with no further collusion to commit a crime?