I wanted to say I that I think it's overrated in terms of its position on HN, but rather than criticize side issues of it, which often point to something being a weak article in general, I probably should have just said exactly what I don't like about it as a whole. So I'll do that.
I think the headline is problematic because it suggests the raw photos aren't very good and thus need processing, however the raw data isn't something the camera makers intend to be put forth as a photo, and the data is intended to be processed right from the start. The data of course can be presented in as images but that serves as visualizations of the data rather than the source image or photo. Wikipedia does it a lot more justice. https://en.wikipedia.org/wiki/Raw_image_format If articles like OP's catch on, camera makers might be incentivized to game the sensors so their output makes more sense to the general public, and that would be inefficient, so the proper context should be given, which this "unprocessed photo" article doesn't do in my opinion.
> I think the headline is problematic because it suggests the raw photos aren't very good and thus need processing
That’s not how I read either the headline or the article at all. I read it as “this is a ‘raw photo’ fresh off your camera sensor, and this is everything your camera does behind the scenes to make that into something that we as humans recognize as a photo of something.” No judgements or implications that the raw photo is somehow wrong and something manufacturers should eliminate or “game”
There are also no citations, and it has this phrase "This website is not licensed for ML/LLM training or content creation." Yeah right, that's like the privacy notice posts people make to facebook from time to time that contradict the terms of service https://knowyourmeme.com/memes/facebook-privacy-notices
> We currently do not support unprivileged use case (same as BPF). Basically, Rex extensions are expected to be loaded by privileged context only.
As I understand it, in privileged context would be one where one is also be able to load new kernel modules, that also don't have any limitations, although I suppose the system could be configured otherwise as well for some reasons.
So this is like a more convenient way to inject kernel code at runtime than kernel modules or eBPF modules are, with some associated downsides (such as being less safe than eBPF; the question about non-termination seems apt at the end of the thread). It doesn't seem like they are targeting to actually put this into mainstream kernel, and I doubt it could really happen anyway..
The experiments of AI agents sending emails to grown-ups are good I think – AIs are doing much more dangerous stuff like these AI Police Reports. I don't think making a fuss over every agent-sent email is going to cause other AI incursion into our society to slow down. The Police Report writer is a non-human partially autonomous participant like a K9 officer. It's wishful thinking that AIs aren't going to be set loose doing jobs. The cat is out of the bag.
Turing test type stuff. Having agents try to participate in the overall social community, by sending a cold email. The reason I think it's good is because with how AI is increasingly used a lot of different sorts of research is needed. A citizen could have their police report written by an AI without them opting in, just like Rob Pike could get a cold email.
> Benchmarks are useful, but they often completely miss out on a lot of real-world factors (e.g., long horizon, multiple agents interacting, interfacing with real-world systems in all their complexity, non-nicely-scoped goals, computer use, etc). They also generally don’t give us any understanding of agent proclivities (what they decide to do) when pursuing goals, or when given the freedom to choose their own goal to pursue.
I'd like to see Rob Pike address this, however, based on what he said about LLMs he might reject it before then (getting off the usefulness train as in getting of the "doom train" in regards to AI safety)
Cincinatti has the real Kentucky across the river. I wonder what that's like. (The reverse could be said as well, and I haven't been to Cincinatti but I'd like to go some day, partly because I saw the Kaufman movie Anomalisa which is set there, and I made a point to check out Schenectady from another Kaufman movie and I was glad I did.)
I was lucky enough to be the only person who showed up for my tour slot, so the guide and I had a lot of time to talk about the art deco, the history, etc.
Oddly enough, the Kentucky side of Cincinnati is nicer than the OH side, at least near the river.
It's worth checking out, but I found downtown Cinci weird. It was completely lifeless at night. Haven't been in 15 years or so, so it's entirely possible that's changed now. Other than that, it's a neat area. Make sure to checkout Jungle Jim's, it's a unique enough store worth a visit.
I was surprised to see something was in text today, until I remembered knowing it at some point - the .har format. Looking at simonw's Claude-generated script [1] to investigate AI agent sent emails [2] by extracting .har archives, I saw that it uses base64 for binary and JSON strings for text.
It might be a good bet to bet on text, but it feels inefficient a lot of the time, especially in cases like this where all sorts of files are stored in JSON documents.
Someone jokingly referred to Glendale, AZ as "Glentucky" once in my presence. I don't remember exactly when and where, I think I remember who, but I am pretty sure it was "Glentucky" even though "Glenducky" would have also worked. :D
reply