refulgentis

5 ms·

61.
by refulgentis 3mo ago
Billions of AI capex for a model that’s consistently forgettable and space computers, presumably data centers in space, are funding SpaceX? Isn’t the opposite? SpaceX has an extremely profitable launch program and a profitable Starlink. It’
62.
by refulgentis 3mo ago
I really do appreciate the effort but the data doesn’t reflect current conditions, and it’s falsified given it’s been virtually closed for months, certainly the same as 30% throughout that is the models default parameter, and we didn’t see
63.
by refulgentis 3mo ago
It’s fine to have it as a normal word, it might even be one! Hehe it is to me at least :) However, I believe OP is right, it’s un…grammatical? -parasable? in the headline? There’s no…referent? More plain spoken :) - there’s nothing indicati
64.
by refulgentis 3mo ago
If we’re prioritizing accuracy: no, that’s not what happened - the blog post about it found by guessing URLs, no access to it was obtained by guessing URLs. Similarly, as long as I’m under the assumption we are prioritizing accuracy: it is
65.
by refulgentis 3mo ago
I didn’t participate in the discussion yesterday because I find it implausible Fable was available long enough (3-4 weeks cumulative?) to get data and train and have it fundamentally affect. But I don’t grok training enough to know that’s
66.
by refulgentis 3mo ago
Hehe, cheers, it really is funny & odd habit I have (usually when I'm in "everyone is wrong!" mode, haven't bothered to argue that, and see someone else arguing it :p)
67.
by refulgentis 3mo ago
You’re absolutely right and it’s heartening to see. I maintain a client with ~every provider you can think of and llama.cpp and it was really tiring the last few days to see people laundering other stuff through Kimi and Qwen. They’re not e
68.
by refulgentis 3mo ago
I did, his argument is that we've already discovered ~50% of all bugs discoverable by LLMs. I'm treading lightly after you said "did you read it" to OP, I do believe we both understand that argument isn't nearly a
69.
by refulgentis 3mo ago
Fable is by Anthropic, and this is too expensive, GLM 5.2 is roughly the same quality at a much cheaper price. (I mantain a client with llama.cpp and 101 models across 14 companies by http)
70.
by refulgentis 3mo ago
There’s a good eval floating around somewhere and tl;dr they’re awesome but the benchmarks are cooked, you’re better off with Qwen 8B Q4 than 27B 1b or ternary. Thanks for being skeptical, I maintain a llama.cpp-based client and it’s frustr
71.
by refulgentis 3mo ago
Fascinating! This is relative time in a continuously processing voice model, here, they're using an LLM with absolute time.
72.
by refulgentis 3mo ago
No, however, if they have the ability to get the current time, they obey constraints like these in a way a model a year ago didn't.
73.
by refulgentis 3mo ago
"Early computer scientists were so optimistic. They beleives with a few kh of ram and a mhz of cpu they could do anything." -- this isn't true, much less the stuff layered on top (conciousness!?)
74.
by refulgentis 3mo ago
Thank you! I'm always finding new rules, even 16 years in :) Much appreciated when it's pointed out.
75.
by refulgentis 3mo ago
That’s fair: The most fair-to-submitter reading of your comment is my concerns are unfounded. If you didn’t intend that, that’s fine. For moderators, I’d suggest that for the community, it’s spam. That’s one example, there are many more lik
76.
by refulgentis 3mo ago
It attracts negative energy via unanswerable questions and downvotes if you assert it is, alas.
77.
by refulgentis 3mo ago
Flagged, lots of unexplained grandiose thinking like “k and q made the vector the unit of thought.” K and q are unexplained and unlinked and “unit vector of thought” is pseudoscientific language Extremely likely to be AI, though I’m not sur
78.
by refulgentis 4mo ago
Yes, support agents never escalate anything, they are mendacious slaves.
79.
by refulgentis 4mo ago
Looks like crap, is it the format?
80.
by refulgentis 4mo ago
Absolutely not.
81.
by refulgentis 4mo ago
Older I get, more it seems it’s possible to ping pong between rewrites for good reasons (ex. here, metric maxes but I find it hard to believe VBR and not-48 kHz are silly things and not worth investing it)
82.
by refulgentis 4mo ago
Thank you for sharing this, I’ve had some disquiet around the release blogs…something felt cherry-picked and I didn’t know there was a 3rd party source for evals, that settles it easily - like you said they can be iffy but there’s a clear e
83.
by refulgentis 4mo ago
OCR is off the charts good on every metric you can think of. LLMs are a near-afterthought at this point if you don’t have data residency requirements. I love them and they’re slightly under rated, their models are consistently well-trained
84.
by refulgentis 4mo ago
The blog post itself is wishcasting, but it’s missing the point completely once we strip .NET. .NET has been around since I was at least 11, I’m 38 now. I don’t know why it’d be considered optional, the backwards compatibility story disappe
85.
by refulgentis 4mo ago
We’re being silly and spamming Slashdot spam comments :p (“imagine a Beowulf cluster” and “first post?”)
86.
by refulgentis 4mo ago
I clicked through and browsed the comments and clicked links. It does appear the links are not working as described. Is there something you’re throwing and I’m not catching? :)
87.
by refulgentis 4mo ago
Imagine a Beowulf cluster of these…
88.
by refulgentis 4mo ago
Your position is “he made it happen by saying he didn’t want it to happen” with maybe a side order of “when he knows Mr. Trump doesn’t like him” I posit that these ideas are common, and come from a place of Mr. Trump is more or less a norma
89.
by refulgentis 4mo ago
Source for this? :)
90.
by refulgentis 4mo ago
No, it’s not leopards ate my face / irony / comeuppance, because that would involve regulation. I understand it’s very satisfying if you wanted Anthropic “punished” for asking for real regulation to see this. I can’t deny there

More ›