Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
nullc
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
10 ms
·
181.
▲
by
nullc
3mo ago
Fair enough!
182.
▲
by
nullc
3mo ago
Just running cpu only w/ Q6 on 9684X I get about 1tok/s ... also still get about 1tok/s/stream when running 16 in parallel.
183.
▲
by
nullc
3mo ago
In the case of the closed models too... Claude would happily tell you it was deepseek-v3 if you asked in chinese until it caught public attention and they papered over it.
184.
▲
by
nullc
4mo ago
> This is important because local llm rarely has parallel streams to batch together. I think most people using agent-like usage could easily run any number of parallel streams pretty often, but you run out of vram for multiple KV caches
185.
▲
by
nullc
4mo ago
I'm a little mystified at people taking about qwen 3.6 27b/ gemma 31b being slow in one breath and then saying they're using a 16GB gpu in the next. You do need to use sutable hardware. I get 50tok/s from Qwen 3.6 27b wi
186.
▲
by
nullc
4mo ago
The paternalistic attitude and historical approach to fictional concerns combined with indifference to actual harms is consistent with the creators of the product. Use a different product. I'm getting fed up with the internet due to s
187.
▲
by
nullc
4mo ago
I don't agree. When governments aren't stopping people or diverting resources, people can always organize themselves to solve problems... and in practice they do. The existence of states is in many senses the formalization of that
188.
▲
by
nullc
4mo ago
Winner winner. Every piece of data a state forcefully collects and retains should be strictly necessary for an important function and balanced carefully against worst case misuse.
189.
▲
by
nullc
4mo ago
Do you think the only, or even primary, thing that determines if a person is marginalized is the color of their skin? -- even taking that you're talking about a POOR neighborhood as a prior?
190.
▲
by
nullc
4mo ago
> but what I think is much more likely My mind went straight to using the AI to write a statement and the AI made stuff up, which would be a nearly guaranteed outcome from using existing LLMs for that task, and it's exactly the sort
191.
▲
by
nullc
4mo ago
Replying to a dead comment that demanded an example, for example, Mao's mass killing of somewhere on the order of 30-40 million people famine (in addition to the million straight up murdered in the cultural revolution) created as a re
192.
▲
by
nullc
4mo ago
I wish the torrent would come before the announcement. Doing it the other way is playing with fire.
193.
▲
by
nullc
4mo ago
> In my ideal the state should have visibility into the shape of the people present so that we can make good decisions about our combined organization. That ideal became tantamount to enabling genocide when the US government breached the
194.
▲
by
nullc
4mo ago
It's hardly even that-- he dumped the remaining customer assets to briefly crash the price of Bitcoin... causing an understatement in what he took from people.
195.
▲
by
nullc
4mo ago
RMS will happily tell you that he'd trade enforcability of the GPL for the non-existence of copyright.
196.
▲
by
nullc
4mo ago
Some of them believe they are building God, and if they can get there first with their God, they can build it in their image and commandeer the free choice of the rest of humanity by force to ensure there will be no God but their God. I wis
197.
▲
by
nullc
4mo ago
what a glorious time to be a plaintiff attorney, subponeas for ai transcripts left and right.
198.
▲
by
nullc
4mo ago
Has anyone evaluated any diffusion LLMs for error spotting? E.g. run your normal autoregressive LLMs (with MTP whatever, as you like), then run a single diffusion pass over the result, and observe any tokens that diffusion thinks are unlik
199.
▲
by
nullc
4mo ago
> "YOLO" is not a reasonable answer here. Yes it is. (1) Ordinary people were able to do these things pre AI-- with some effort into study for sure. (2) The cat is already out of the bag, open models can already help with thes
200.
▲
by
nullc
4mo ago
These chicken littles have lied far too many times in far too extreme ways in their desperate attempts to obtain a monopoly by regulation. The remaining 'worthwhile message' would be that they have deleted their models and are dis
201.
▲
by
nullc
4mo ago
The FTC settlement with GM allows GM to sell precise location as long as it's anonymized by attaching it to anonymous identifiers rather than personal info. It also allows non-precise location (e.g. zipcode/census-block) attached
202.
▲
by
nullc
4mo ago
Sorry to be a debbie downer, but this reads like LLM slop rather than engineering work. I don't just mean the language on the page-- although that too (not an X it's a Y, over and over again)-- but the absence of the artifacts o
203.
▲
by
nullc
4mo ago
I think that's an extremely well done response on his part.
204.
▲
by
nullc
4mo ago
If you don't follow up with a pass of injecting bugs and validating that the tests fail in the presence of bugs... then you've only confirmed that the tests can pass and they may be substantially useless.
205.
▲
by
nullc
4mo ago
Wrong metric-- the person caught would have almost certainly been caught absent it, making it easy to overstate the benefit. When someone with access-- potentially LEO but the access set is much larger-- uses the data to stalk and harass so
206.
▲
by
nullc
4mo ago
Don't type it yourself, automate the abuse.
207.
▲
by
nullc
4mo ago
Claude's fictional inspiration issue is more general than just how it behaves when given the freedom to act. There is an ongoing issue with nutters going to claude with conspiracy theory premises and the AI just riffs along with the t
208.
▲
by
nullc
4mo ago
Classically the training process is entirely about imitation and not at all about reasoning. Imagine you're training an LLM (a text predictor) on a corpus consisting of "The AI agent was switched on and then ran the command {takeo
209.
▲
by
nullc
5mo ago
LLMs simulate human language as it is used by humans. The usage by humans demonstrates evidence of empathy, motivations, etc. So we should expect LLMs to exhibit similar traits to the extent that it hasn't been carefully avoided in th
210.
▲
by
nullc
5mo ago
It's not his employer that has the rights-- it's the publisher which at no point paid for the research. As an American tax payer I funded the poster's research. And yet if I want to read about it I have to pay a foreign pr
More ›