arjie

11 ms·

241.
by arjie 2mo ago
Haha! Incredible. What a scam! Too good to be true, I suppose. I never used it for anything meaningful because everything meaningful is also tragically sensitive so I suppose I didn't notice.
242.
by arjie 2mo ago
Every time I see it I think of those YouTube videos people used to make: Ben Shapiro OWNS feminist with FACTS and LOGIC. Definitely sets the bozo bit.
243.
by arjie 2mo ago
Yunwu.ai was a nice way to get ultra-cheap tokens that someone will probably spy on you with, but they no longer offer US models. At first I thought it was so people could steal the traces, but now I wonder if this isn't just launderin
244.
by arjie 2mo ago
Inference is profitable. Even open models can be profitable. Obviously your cost of capital is the thing that influences that the most.
245.
by arjie 2mo ago
It's obviously not in their interest. Disclosing would only work as marketing for their IPO and it's the most anticipated IPO in history and needs no marketing. There is like zero positive side to disclosing audited financial stat
246.
by arjie 2mo ago
I wouldn't do it without pgbouncer. Asking for trouble when one day connections exceeds. It's just that you start with "oh I'll manage the pool from my app" and then you're stuck with either putting things into
247.
by arjie 2mo ago
Diesel, Ethanol, Pyrolysis? Haha, odd. I know this is supposed to be “not allowed in government documents” and so on but I think it would be useful to have a list of “words that are not permitted to be used in the US”. It could be a conveni
248.
by arjie 2mo ago
Interesting. My dad wears his on the back of the arm (posterior) not the inside (medial) - and when I asked him he pointed to the official instructions which do actually have it there https://www.freestyle.abbott/content&#x2
249.
by arjie 2mo ago
I see. The way I spoke made it seem like I was talking about the Copenhagen wheel with the phrase “this particular one”. I was talking about the OP device. Obviously since Mission and SVN is in the US I would have assumed it was not the CW
250.
by arjie 2mo ago
Haha, I like all these adapter designs though my favourite was the Copenhagen Wheel[0]. I saw someone ride one near Mission and S Van Ness and loved it. Sleek device. This particular one in the OP is definitely built for safe societies, rig
251.
by arjie 2mo ago
Unfortunately the reverse is also true: polderization and exposing the peat to the air reverses that sink process. Single digit percentage of Dutch greenhouse emissions come from this. Fortunately, they are depolderizing and hopefully much
252.
by arjie 2mo ago
Most of us aren’t deploying these in general purpose use-cases. E.g. I use Qwen mostly for vision in my personal assistant. I have an eval set for that. Pretty much each of my use cases has a pre-computed problem set. Likewise I have a tool
253.
by arjie 2mo ago
A similar story around this is the controversy around various Mumbai “encounter specialists” (a violent incident between police and gangsters is an “encounter”) and whether they were effectively hitmen for some faction of the local mob. Alw
254.
by arjie 2mo ago
Interesting. Here's the section of the EU Act that mandates this: > Providers of AI systems, including general-purpose AI systems, generating synthetic audio, image, video or text content, shall ensure that the outputs of the AI sy
255.
by arjie 2mo ago
Same, with a multimodal DSv4 Flash I would just stop paying attention to things. Very smart, and at 260 tok/s it's too fast to care about anything else. If you ever graft something like that I would love to hear about it. Yes, I h
256.
by arjie 2mo ago
I use the Qwens as a vision model for my DeepSeek V4 Flashes to handle. But the Qwens run on old RTX A6000 Ampere. Does anyone know if there's any news about INT4/AWQ quants for the RTX A6000?
257.
by arjie 2mo ago
That’s fair. You should be able to play with an idea without it being perfect. All right then. I think industry level collusion is the natural thing that arises from this with regulatory capture. But it’s fine. No need to be overly negative
258.
by arjie 2mo ago
I think the thought experiment of achieving marginal cost per job higher with human than machine for all jobs is best explored with the pathway there. Machines will not come into existence in a blink. Real-world concerns like manufacturing
259.
by arjie 2mo ago
Prompt expansion of input to extra categories makes sense if your embedding isn’t working well. But on its own, why use the LLM at all? I think you could have demonstrated the original step first and then shown that it’s useful.
260.
by arjie 2mo ago
Market economics must be a truly unstable equilibrium in a cognitive sense. People constantly reinvent centrally planned societies where jobs of need are chosen by the state. Inevitably, in their conception these jobs are chosen by the util
261.
by arjie 2mo ago
Yes, that's true. There are many tasks you cannot do this on, but it's far beyond reading other people's broad benchmarks IMHO. e.g. there are tasks where Gemma is great, but on my camera watching, Qwen 3.6's MoE far exc
262.
by arjie 2mo ago
I mean "since the actual environmentalists nonetheless act as rearguard for the others, the distinction is faint". i.e. it's not interesting to me that person A wears mask M1 and suggests viable alternatives and then puts on
263.
by arjie 2mo ago
In the US, it's pretty easy. My wife used to live in the Tenderloin for $2400/mo and her place was nicer than this because she was good at putting things together. I don't happen to have a photo of the whole studio handy, but
264.
by arjie 2mo ago
Yeah, because all those employees are now doing much better paying jobs elsewhere instead of a low paying job like cleaning a dining room. The dining room at a McDonald's functions just as well at the 80% dirtiness as it did at 10% dir
265.
by arjie 2mo ago
It's funny to me that entire companies exist that have a fraction of storage that I personally have in a datacenter on flash let alone in total. 50 TiB is so little. It's like $7k annually on R2 and under a thousand bucks on S3 Gl
266.
by arjie 2mo ago
I went to Glacier National Park a while ago and stood before a sign that said "The glaciers you see will be gone by year 20xx" or something like that. The year was 20xx and I was looking right at the glaciers and they were decided
267.
by arjie 2mo ago
Building ad-hoc evals is trivial these days. And you can place an LLM judge in front to disambiguate between output. e.g. I have Frigate gating a video feed so that an LLM can watch our home cameras and the labeling task takes a few minutes
268.
by arjie 2mo ago
If warrant requests are guaranteed accepted or rejected in t<=15 mins, then I think I support the idea that they should be required with the functionality that they are auto approved at some t+5 deadline. Just to clarify, the timelines y
269.
by arjie 2mo ago
I live in San Francisco. Paid like $80 or something back then but I drank the whole thing over the next few months after finding out there was nothing dramatic going on. $100 seems about what I'd expect. Is that historically a high pri
270.
by arjie 2mo ago
Just for my own epistemic hygiene, would you mind clarifying your familiarity with applying for and receiving a similar warrant and the duration you envision when you say "wouldn't take a long time"? Very reasonable not to, o

More ›