I remember it being somewhat common for people to make forum posts consisting entirely of a joke image. However, they weren’t called memes at the time as the word had yet to be popularized.

Job market statistics are actually probably a strong point for the AI overview. I just Googled 'us job market last month' and got an AI overview that accurately summarized a New York Times article for qualitative information ("surprisingly strong 115,000 jobs", "no-hire, no-fire"), followed by accurately summarizing the official Bureau of Labor Statistics website for raw stats, followed by some other stuff I didn't check. Not everyone would prefer The New York Times' take, but the citation prominently displays their name and logo, so you can tell what you're getting.

Weak points are when the topic is obscure enough that the AI overview conflates two different things or overgeneralizes, or trusts the wrong sources.

> Decline to buy: property stays with bank (auction abstracted out)

Ignoring an entire game mechanic is really stretching the definition of “abstracted out”…

Also, at the bottom it defines a “Liveness: someone eventually wins” property which I believe cannot be proven. Monopoly doesn’t have any rules forcing the game to end eventually. There is only a probabilistic guarantee, and even that only applies if the players are trying to win; if the players are conspiring to prevent the game from ending then they’re unlikely to fail.

> Note that nothing in this objective constrains the NLA explanation z to be human-readable, or even to bear any semantic relation to the content of [the activation].

The objective could be optimized even if the verbalizer and reconstructor made up their own “language” to represent the activations, that was not human-readable at all.

To point the model in the right direction, they start out by training on guessed internal thinking:

> we ask Opus to imagine the internal processing of a hypothetical language model reading it.

…before switching to training on the real objective.

Furthermore, the verbalizer and reconstructor models are both initialized from LLMs themselves, and given a prompt instructing them on the task, so they are predisposed to write something that looks like an explanation.

But during training, they could still drift away from these explanations toward a made-up language – either one that overtly looks like gibberish, or one that looks like English but encodes the information in a way that’s unrelated to the meaning of the words.

The fascinating thing is that empirically, they don’t, at least to a significant extent. The researchers verify this by correlating the generated explanations with ground truth revealed in other ways. They also try rewording the explanations (which deserves the semantic meaning but would disturb any encoding that’s unrelated to meaning), and find that the reconstructor can still reconstruct activations.

On the other hand, their downstream result is not very impressive:

> An auditor equipped with NLAs successfully uncovered the target model’s hidden motivation between 12% and 15% of the time

That is apparently better than existing techniques, but still a rather low percentage.

Another interesting point: The LLMs used to initialize the verbalizer and reconstructor are stated to have the “same architecture” as the LLM being analyzed (it doesn’t say “same model” so I imagine it’s a smaller version?). The researchers probably think this architectural similarity might give the models some built-in insight about the target model’s thinking that can be unlocked through training. Does it really though? As far as I can see they don’t run any tests using a different architecture, so there’s no way to know.

However, the shell could theoretically chunk long cd arguments into multiple calls to chdir(), splitting on slashes. I believe this would be fully semantically correct: you are not losing any atomicity guarantees because the kernel doesn’t provide such guarantees in the first place for lookups involving multiple path components. I’m not surprised that bash doesn’t bother implementing this, and I don’t know if I’d call that an “arbitrary limitation” on bash’s part (as opposed to a lack of workaround for another component’s arbitrary limitation). But it would be possible.

The way the OP used the 'not X, but Y' pattern, the 'X' and 'Y' are two clear, specific, and (most importantly) distinct things, as opposed to stereotypical LLM usage where they're vague characterizations or metaphors. And there's a reason to emphasize that it's not X, because the Slop Cop website implicitly suggests that it is X.

https://github.com/apple-oss-distributions/xnu/blame/f6217f8...

https://github.com/apple-oss-distributions/xnu/blame/f6217f8...

But yes, “assert” in most languages is debug-only.

Out of the two claims, the only one that made it to appeals court was about whether it was fair use for Bleem to use screenshots of PS1 games to advertise its emulator (which was compatible with those games). The Ninth Circuit decided it was. But that's not relevant here.

The other claim was more relevant, as it was an unfair competition claim that apparently had something to do with Bleem's reimplementation of the PS1 BIOS. But the district court's record of the case doesn't seem to be available online, and the information I was able to find online was vague, so I don't know what exactly the facts or legal arguments were on that claim. Without an appeal it also doesn't set precedent.

If there were a lawsuit over OpenTTD, it would probably be for copyright infringement rather than unfair competition, and it would probably focus more on fair use and copyrightability. For fair use, it matters how much something is functional versus creative. The PS1 BIOS is relatively functional, but a game design and implementation are highly creative. On the other hand, despite being creative, game mechanics by themselves are not copyrightable. So it might come down to the extent to which OpenTTD's code was based on the reverse-engineered original code, as opposed to being a truly from-scratch reimplementation of the same mechanics. Visual appearance would also be relevant. Oracle v. Google would be an important precedent.