I tried this with the original comment in the thread. Guaranteed to not be in the corpus, references a few terms that also wouldn't be in the corpus (Claude Fable), and long enough to be more than a sentence or two while short enough to compare in a discussion like this.
I did this with entirely local models I have sitting around on my laptop. Minimax M2.7 at a 3 bit quant with 8 bit quantized KV cache for English -> French, Gemma 4 31B QAT (4 bit quant) MTP for French -> English.
It's perfectly readable, but there are a few places where the phrasing is a bit more awkward after the double translation ("auditing" to "revision" in particular is a bit off). Gemma did comment on not knowing what Claude Fable was in its thought process: "The author compares Ellsworth's translation with one produced by "Claude Fable" (likely a misspelling of "Claude" or a specific version of Claude)."
Here's the double translation:
"I have no doubt that a writer is better at translating than AI, but I must say that AI translation has become so good that I'm not sure how much longer the profession of translation will exist—or rather, it may become more a matter of revision.
"For example, I just read Lawrence Ellsworth's translation of The Three Musketeers, which I enjoyed immensely. I neither speak nor read French, but from what I understand, Ellsworth's translation is considered one of the most faithful translations of the work.
"Out of curiosity, I asked Claude Fable to translate the original French version of The Three Musketeers; I asked it to translate faithfully, but also to try to maintain the same playful tone as the original and to censor nothing.
"Once it was finished, I didn't read the entire result, but I compared a few individual chapters between Ellsworth's translation and Fable's.
"They were honestly remarkably similar. As far as I can tell, nothing was substantially different between Ellsworth's translation and Fable's. I think the prose in Ellsworth's translation was slightly better, but Fable's was actually perfectly readable. Again, I don't speak French, so I can't say for certain, but I don't believe I would have had a significantly different experience if I had read Fable's version instead of Ellsworth's.
"It is possible (and probable) that this is partly a self-fulfilling prophecy; Fable may have been trained using Ellsworth's translation and can therefore draw directly from it. Unfortunately, since I don't speak any language other than English, there is a sort of vicious circle: the only way to compare the fidelity of a translation is to compare it to other translations, but if other translations already exist, that will likely influence the results, and if a translation doesn't exist yet, I have no way of verifying it.
"I am going to continue reading Ellsworth's translations for the following stories simply because it feels more canonical to me, and as I said, I think the prose was slightly better."
Phenomenal author indeed. "Capital", a novel about the financial crisis and bankers on the hedonic treadmill, is wonderful too. Too bad either Lanchester or his publisher has beef with Amazon, most of his work is not available as an ebook.
To anyone for whom Houston is a bit too far of a drive (plus the fact that the museum now seems closed), but Cincinnati is not: Union Terminal is another beautiful example of art deco.
Fellow Union Terminal aficionado here. Also highly recommend to any art deco architecture lovers.
Even if you're not into art deco architecture it's worth a visit. The Omnimax theatre is worth the trip alone, but the museums are great too.
@macintux - It's hard to get on the tour because they sell out, and they don't allow photography, but the high steel tour, which includes walking up to the top of the inside of the half dome, is great. There's some duplication with the behind the scenes tour but it's still worth it. It's probably second to the (now discontinued for liability reasons) Cincinnati subway tour in terms of seeing cool hidden stuff in the Cincinnati area.
"The boy who cried wolf" is a story about false positives, so if that's what you want to avoid then you want to get close to 100% specificity, and accept that there are many things that the tool will not catch. If, as you propose, the tool would mainly be used to create a low confidence list of potential problems that will be further reviewed by a human, then casting a wide net and calibrating for high sensitivity instead does make sense.
The idea is to minimize the false positives "the boy who cried wolf" at the same time mitigate, or better eliminate false negatives. The main reason is that based on the physician in-the-loop, the system can be optimized for sensitivity but can be relaxed for specificity. Of course if can get both 100% sensitivity and specificity it will be great, but in life there's always a trade-off, c'est-la-vie.
In our novel ECG based CVD detection system we can get 100% sensitivity for both arrhythmia and ischemia, with inter-patient validation, not the biased intra-patient as commonly reported in literature even in some reputable conferences/journals. Specificity is still high around 90% but not yet 100% as in sensitivity but due to the physician-in-the-loop approach, which is a diagnostic requirement in the current practice of medicine, this should not be an issue.
Before Netflix was a thing, I sometimes tried to have conversations with people about "gee, it's a bit annoying that my only options to watch a movie is to buy an expensive dvd that I will watch once, or to pirate it" and the most common response was complete befuddlement, they simply could not comprehend that someone might not want to pirate things if they could, they could not comprehend that besides being illegal it was also just... wrong. Not absolutely evil, for sure, but still something that maybe you might want to avoid doing. Now that you can just pay 10-20 euro for a streaming service, most of them have switched over, so, yeah, service does matter, but a lack of risk or consequences on the one hand and vague notions about actors and directors (and soccer players) already being rich enough as it is, were enough to convince very many people that piracy was a victimless crime.
> Now that you can just pay 10-20 euro for a streaming service,
The nice thing about piracy is that you can find what you want immediately. You don't have to go to an aggregator site to find out where it's available, and then log on to the streaming platform site to find that the aggregator site is lagging the real availability, or find that certain content isn't available in your country, or that the content is available but only on the special extra++ cost plan instead of the basic plan.
If you want to watch content legally, the workflow looks like this:
Search content -> go to aggregator site -> select streaming site -> enter electronic contact and payment info and physical address (for payment) -> confirm email account -> watch content -> dig around on site to find deliberately hidden unsubscribe workflow -> pass all the "are you sure you want to leave" screens -> monitor your card payment the next month to make sure you actually cancelled
The illegal workflow looks like this:
Search content -> click 1-3 sketchy sites, closing 15 pop up ads -> watch content -> forget about it
I strongly believe the fact that media companies struggle to accept payments worldwide and region-lock their content when you do pay is why their services ultimately suck for customers.
Eg. for my HBO GO subscription provided by my cable operator to continue working, I had to disable load balancing/failover between my other ISP for HBO addresses at home or it'd just stop working when it detects I've been switched to a different network. And then you travel and can't access it anymore either. It is completely bonkers.
As a sibling comment said, Netflix won (at that point) because they made service easy and converted a bunch of customers over.
You just have to go along with the idea that skin provides no indication of meatiness and that the two aliens are Ford Prefect types, then the short film lands just fine.
I guess. It's still hard to mesh with the idea they don't believe these humans flap their meat at each other, or that they do not communicate exclusively via radio signals.
It doesn't match my idea that these are two energy/mechanical beings discussing a faraway planet from their spaceship or whatever, talking theory without actually seeing the beings they are discussing.
You've never encountered, say, a baffling code bug that couldn't possibly be caused by X, spent a day on it, and found out it turns out to be caused by X?
More seriously, what you describe is partly the short story. The short film adaptation doesn't quite work for me, for the reasons I explained in other comments.
The AITA comparison seems apt insofar as chatbots function as a second opinion. You're consciously or subconsciously looking for an outside perspective that might differ from that of your friends, provided to you by a computer that doesn't need to care about your feelings, unlike a friend. If the chatbot ends up mimicking what (not very close) friends do, you might falsely conclude that two very different kinds of sources have converged on the same answer, whereas you are really just getting two flavors of the same diplomatic interaction.
My subjective impression is that 5 years ago AITA was actually quite wholesome and the top comments tended to be insightful. The shift towards "set boundaries, always choose yourself, you don't owe anybody anything" seems fairly recent.
I get the impression that most desktop users enable zram or zswap to get a little bit more out of their RAM but there is never any real worry about OOM, not regularly anyway, so then (according to the principles laid out in the article) it shouldn't matter much.
On my workstation, I run statistical simulations in R which can be wasteful with memory and cause a lot of transient memory pressure, and for that scenario I do like that zswap works alongside regular swap. Especially when combined with the advice from https://makedebianfunagainandlearnhowtodoothercoolstufftoo.c... to wake up kswapd early, it really does seem to make a difference.
It doesn't really need any config on most distros, no.
That said, if you want it to behave at its best when OOM, it does help to tweak vm.swappiness, vm.watermark_scale_factor, vm.min_free_kbytes, vm.page-cluster and a couple of other parameters.
reply