ls612

8 ms·

151.
by ls612 5mo ago
I’m really wishing I had overbuilt my NAS last year. As it stands I feel lucky to have even built it at all given I bought all the parts in the last week of September. My 4090 and 12900k are gonna have to last till 2029 at this rate won’t t
152.
by ls612 5mo ago
Humans have a hard time judging how deep water is too! Turns out neither Lidar nor vision/cameras have the right ability to sense water depth.
153.
by ls612 5mo ago
Do you still need a wacky backend to run them locally or does LM Studio make it easy nowadays? Last I use a local diffusion model was late 2022.
154.
by ls612 5mo ago
Is SDXL still the best local image model all these years later? Damn, that’s sad…
155.
by ls612 5mo ago
Even not assuming Blackwell inference the $3.50/hr price is likely close to the marginal cost. The Deepseek R0 model is a little more than a third of the size of V4 and cost around $1/Mtok to serve at scale based on deepseek'
156.
by ls612 5mo ago
Yes it is more efficient in $/tok to run at scale than to run just for yourself. Everyone selling Deepseek V4 inference is selling an undifferentiated good. They have run the numbers on how much it costs and are competing against a do
157.
by ls612 5mo ago
The solution I use for remote access is tailnet plus exit nodes. It also is probably more secure as the attack surface is wireguard + the tailscale control plane rather than an entire internet facing Plex server.
158.
by ls612 5mo ago
Claude has gotten good in the past month or two at recognizing when it might need to search the web for updated info rather than saying that it has no idea what I'm talking about or making stuff up.
159.
by ls612 5mo ago
Anyone can host Deepseek V4 on rented GPUs and sell inference on it. Price will very quickly converge to the marginal cost of inference. This is as close to a pure commodity as it gets in the AI space so competitive market economics will pu
160.
by ls612 5mo ago
He lost the lawsuit on a legal technicality about the statute of limitations not on substantive grounds.
161.
by ls612 5mo ago
10 tok/s is around the borderline of interactive being good. I did the math and it is mostly bottlenecked by memory bandwidth, so in the future I can expect to run a similarly sized model on my 4090 once it gets retired from gaming ser
162.
by ls612 5mo ago
We went through this whole song and dance in 2014. Unless it has some really unlucky novel mutations it won’t spread well outside the tropics.
163.
by ls612 5mo ago
I run it on my 4 year old MBP and get 10 tok/s. With the RAM shortage buying anything new today is a nightmare but anyone with a reasonably modern Mac could run it at q6 probably. It is mostly a toy as 4o models weren’t really suitable
164.
by ls612 5mo ago
As good as today’s frontier. Gemma 4 today is roughly equivalent to the frontier a year and a half ago at gpt 4o tier.
165.
by ls612 5mo ago
Yes in the long term. Sometimes they will get surprised by unexpected events (Covid) but you can’t blame them when their tourism and hospitality numbers are wildly wrong when that happens.
166.
by ls612 5mo ago
This deserves a lifetime ban for being a bad researcher obviously. (/s)
167.
by ls612 5mo ago
It sounds like even copy/paste errors from Google Scholar would qualify for the lifetime ban. Let he who is without sin…
168.
by ls612 5mo ago
The estimates I've seen are that running inference at scale on a Deepseek V3 sized model (so 700B parameters) costs roughly $0.70/mtok or so given current H100 rental costs. Sonnet charges $15/mtok on the API so the delta bet
169.
by ls612 5mo ago
For the ruling class. Cattle classes are not people in their eyes.
170.
by ls612 5mo ago
This is how you get squished like an ant by an ambitious federal prosecutor using the CFAA.
171.
by ls612 5mo ago
It is easy to change the system prompt to make the AI talk with a different voice. It is remarkably hard (at least for Claude, I haven't experimented as much with GPT) to get it to not use so many em-dashes like this essay does.
172.
by ls612 5mo ago
This whole piece is AI generated.
173.
by ls612 5mo ago
Place the chinese labs on the entities list. That stops any legitimate company using them and probably makes HF take them down. Sure there will be torrents but the laws for doing business with a sanctioned entity bite much harder than the l
174.
by ls612 5mo ago
But we wouldn’t be. I’m assuming that the US labs retain several months’ lead for at least the next couple of years.
175.
by ls612 5mo ago
If America just banned all chinese models that would wipe out most of the open weights landscape in AI, especially anything close to the frontier. I could easily see that happening if a Mythos tier model comes out of a Chinese lab in early
176.
by ls612 5mo ago
I think you misunderstand the point I'm making. Governments love having this centralized ability to attest hardware and control what software can be run. This is why for instance the EU has really slow-walked and watered down side lo
177.
by ls612 5mo ago
I’m on an M2 Max and get 10 tok/s with Gemma 4 8bit MLX
178.
by ls612 5mo ago
I’m hopeful that in a year or so the models will be good enough to help productively with emulator development and that you will see a similar shift to these PRs that you did with security this spring.
179.
by ls612 5mo ago
My point is that as far as I understand (not a cryptography expert) once you have the mathematical concept of asymmetric cryptography you also have the mathematical concept of a certificate, so you can't have one without the other.
180.
by ls612 5mo ago
When one group says “we don’t want surveillance” and the other group says “we will use surveillance to destroy you” the equilibrium is clear. This is why liberalism will not survive in the 21st century.

More ›