derefr

25 ms·

211.
by derefr 8mo ago
A-frames are often used in snowy climates as a vacation home, park ranger patrol station, or https://en.wikipedia.org/wiki/Wilderness_hut . Such buildings are un-lived-in for much of the year, if not "indefinitely
212.
by derefr 8mo ago
> This can be dangerous in 2 ways. 1. Weight and 2. if it randomly comes all crashing down potentially hitting people. A-frames are steeply-angled enough that the weight mostly loads on itself (i.e. a snow drift builds up against the sid
213.
by derefr 8mo ago
> I also balk at how much extra computing power is needed to generate effects I find no value in and would prefer to disable. I mean, "computing power" in a literal sense maybe, but does that matter if it doesn't translate
214.
by derefr 8mo ago
I spent more than an hour writing the above comment, with my own two human hands, spending real thinking time on inventing some (AFAIK) entirely-novel educational metaphors to contribute something unique to the discussion. And you're g
215.
by derefr 8mo ago
Oldschool telnetd didn’t actually run as root; rather, it just set up a PTY for the incoming socket to talk to, and then fork-exec’ed a /bin/login subprocess to live inside that pty. /bin/login is setuid-root, so it’s “w
216.
by derefr 8mo ago
Exactly. My personal mechanistic understanding of diffusion models is that, "under the hood", the core thing they're doing, at every step and in every layer, is a kind of apophenia — i.e. they recognize patterns/textur
217.
by derefr 8mo ago
> like its composited Like focus stacking, specifically. I’m always surprised when people bother to point out more-subtle flaws in AI images as “tells”, when the “depth-of-field problem” is so easily spotted, and has been there in every
218.
by derefr 8mo ago
The negative impacts of generative AI are most sharply being felt by "creatives" (artists, writers, musicians, etc), and the consumers in those markets. If the OP here is 1. a programmer 2. who works solely with other programmers
219.
by derefr 8mo ago
> That's not how training works - adjusting model weights to memorize a single data item is not going to fly. Apologies; I think I got us all kind of off-track in this comment thread by stretching the definition of the term "fi
220.
by derefr 8mo ago
Doesn't necessarily need to be online . As long as: 1. there's a way to take many transcripts of inference over a period, and convert/distil them together into an incremental-update training dataset (for memory, not for RLHF
221.
by derefr 8mo ago
> And the technical implementation used can't be divorced from everything that surrounded it: Google's place in the industry with regard to search engines, ads, etc. I mean... sure it could have? There could have been an indepe
222.
by derefr 8mo ago
I wonder why they don't just prioritize the ~500 most popular of those content providers that are feeding them sludge articles, and write (AI-generate, even) logic to manually parse and transform said sludge into their format? It'
223.
by derefr 8mo ago
> these responses go beyond role play Are they sure? Did they try prompting the LLM to play a character with defined traits; running through all these tests with the LLM expected to be “in character”; and comparing/contrasting the r
224.
by derefr 8mo ago
Something TFA doesn’t mention, but which I think is actually the most important distinction of all to be making here: If you follow this advice naively, you might try to implement two or more of these other-kind-of-DB simulacra data models
225.
by derefr 8mo ago
You need to look closer; the heart emoji has a flat fill, but a gradient in its outline stroke, from lighter-than-red near the top, to darker-than-red on the bottom.
226.
by derefr 8mo ago
Well, on the linked page, it mentions "strong transcription performance in 13 languages, including [...] Hindi" but with no mention of Bengali. It probably doesn't know a lick of Bengali, and is just trying to snap your words
227.
by derefr 9mo ago
No; if a phone has both a non-removable battery and a baseband modem, then various laws require that modem to be wired directly to that battery (and to the phone's microphone) and to able to be activated in response to a legal wiretap
228.
by derefr 9mo ago
Architects aren't generally brutalists themselves, but rather, brutalist architecture proposals win contracts because their TCO is lower. Facades have maintenance costs; bare concrete just requires power-washing now and then.
229.
by derefr 9mo ago
I don't think the exciting thing here is the technology powering it. This isn't a story about OpenClaw being particularly suited to enabling this use-case, or of higher quality than other agent frameworks. It's just what peop
230.
by derefr 9mo ago
My point was that one of the devices would be your (cold) backup — you'd e.g. get an (ideally passphrase-protectable) smart-card; read off its pubkey; register that pubkey with all your remote systems/services; and then put the
231.
by derefr 9mo ago
In countries with public healthcare + doctor shortages (e.g. Canada), good luck even getting a family doctor, let alone having a request to switch you family doctor "when you already have one!" get taken seriously. Everyone I
232.
by derefr 9mo ago
I think the implication is that you should own multiple client devices capable of SSHing into things, each with their own SSH keypair; and every SSH host you interact with should have multiple of your devices’ keypairs registered to it.
233.
by derefr 9mo ago
I believe that will result in Google locking you out of your Google account, including Gmail, YouTube, any Google Cloud projects, etc.
234.
by derefr 9mo ago
It’d be difficult to use in any automated process, as the judgement for how good one of these renditions is, is very qualitative. You could try to rasterize the SVG and then use an image2text model to describe it, but I suspect it would jus
235.
by derefr 9mo ago
The convenient thing in this case (verification of translation of academic papers from the speaker's native language to English) is that the authors of the paper likely already 1. can read English to some degree, and 2. are highly li
236.
by derefr 9mo ago
I would argue that an LLM is a perfectly sensible tool for structure-preserving machine translation from another language to English. (Where by "another language", you could also also substitute "very poor/non-fluent E
237.
by derefr 9mo ago
> Coming from a software background, this seems bizarre, as if C++ compilers rejected valid programs unless they stuck to easy constructs with obvious assembly implementations. To my understanding, isn’t it more like there being a perfec
238.
by derefr 9mo ago
It occurs to me that "write a C program that [problem description]" is an extremely under-constrained task. People are highly aware that C++ programmers are always using some particular subset of C++; but it's not as obvious
239.
by derefr 9mo ago
I follow the author's line of reasoning, but I think that following it to its logical conclusion would lead not to an `execute_code` primitive, but rather to an assumption that the model's stdout is appending to a (Jupyter, Livebo
240.
by derefr 9mo ago
Sounds like an approach that would also work for ML model weights files — just another kind of multidimensional array with metadata. I wonder what exactly the big multi-model AI companies are doing to optimize model cold-start latency, and

More ›