tmalsburg2

That depends on the country and source of the data. In Germany, pretty much all buildings are tagged with house numbers.

How can it be dense and wordy at the same time? One is the opposite of the other.

It’s good for beginners (also a great reference). However, stuff gets serious pretty fast, and the reader is expected to put in serious time and effort.

Source?

No, I do not think that all lawyers are dumb. What a bizarre thing to say. Why derail an interesting conversation in such an aggressive way?

If this phenomenon is so obvious, why is this story the top post on HN? And why did it take a picture of the mirror scenario to make people aware of this issue? Hell, the article even implies that this an issue only with images of mirrors when that is of course completely false.

Completely valid nitpick.

It’s unlikely, true. But precisely that makes it so dangerous. If such a picture is presented as evidence in a murder case, the possibility that it is telling the wrong story will be discounted and someone may go to prison for the rest of their lives.

Rolling shutter could in principle explain this, but it’s probably too fast to capture these three poses which must be temporally rather far apart.

This effect likely resulted from aggressive noise filtering. You can probably reduce the amount of filtering in the settings.

Imagine such an image being used as evidence is a court case. E.g. showing someone pointing a gun at another person when that actually never happened.

Not clear at all. Ilya leaving would look really bad for OpenAI. Altman needs him to stay, and he keeps the door wide open for that in the statement. Question is: How much is he willing to offer (money and otherwise) to make that happen?

I doubt it. It’s too damn costly computationally.

What you say is correct but it doesn’t explain why the model fails to capture the semantics of “is”.

This is a useful observation, but it doesn’t explain the particular example given in the article.

Why do you think that tokenization is to blame on this particular limitation?

It's the whole point of Emacs that it's malleable and can (should) be tailored to the own personal users needs.

Sorry, but that's again a straw man: No one claimed that those countries never have bad news. Those countries having bad news and the Economist exploiting it also doesn't really seem like a crazy conspiracy theory.

Not surprised. These models are inherently stochastic and will therefore happily say A and ¬A when prompted multiple times. A fundamental difference to how humans operate. (You could say that humans are stochastic, too, but they're stochastic in different ways.)

GPT4 response:

  You can translate the phrase "She said goodbye too many times before" into Latin as "Dixit vale saepe nimis antea".

I asked GPT4 whether it could make the translation shorter to which it responded that Latin was inherently a verbose language, so no.

Nobody said they were a panacea. Don't try to derail a potentially useful conversation.

It is easy to detect what people are looking at. Solved problem. It’s difficult to effectively use such a low—bandwidth signal, and that’s where LLM come in. You need an LLM to anticipate what the user wants to say. An LLM fine-tuned on the user’s needs may have sufficiently low entropy that the small number of bits/s you’re getting from the eye-tracker can reliably identify the intended message.

Solving eye-tracking keyboards is not so much a task for a company with eye-tracking expertise but for one with expertise in large language models.

C is neither minimal nor well-defined, as has been discussed extensively. Random blog post illustrating only some of the issues is here [1]. Just google “undefined behavior c” for a lot more. There are also a bunch of famous rants by Linus Torvalds on his topic.

[1] https://blog.regehr.org/archives/213

> What is "work" in this context?

I'm primarily interested in 1. having all hardware work out of the box, 2. working suspend to disk/RAM, 3. not having to edit `Xorg.conf`, 4. relatively painless upgrades.

Ubuntu gives me all of that and has served me well for 15 years. But their transition to snap forces me to look for something else.

I used Debian as a student when I had time to endlessly edit config files. What's the state today? Is it possible to just use it to get work done, say, on a Thinkpad?

The fact that the author lists her pronouns is probably enough to offend some people here.

Why has this submission been flagged? I see nothing wrong with it, and the discussion here is constructive and insightful.

I don’t think that this list is supposed to be an accurate representation of how people typically use vim. More likely this list was compiled precisely because these uses are somewhat non-standard / non-obvious and therefore worth sharing. Having said that, I do believe that using external tools is much more idiomatic for Vim than Emacs.

You interpret X11 apps as a technical descriptor, but I think it's more a description of the ecosystem in which these apps evolved and whose culture they represent.