You're revoking the attacker's key (that they're using to upload the docs to their own account), this is probably the best option available.
Obviously you have better methods to revoke your own keys.
3,006 karma · joined October 23, 2012
Obviously you have better methods to revoke your own keys.
Optimizing a language for LLM consumption and generation (probably) doesn't mean you want a LLM designing it.
OpenAI? The OpenAI head of research @markchen90 straight up admitted it in a podcast.
https://x.com/petergostev/status/1995744289079656834
"In the last 2 years we've put so much resourcing into, into reasoning and one byproduct of that is you lose a little bit of muscle on pre training and post training." "In the last six months, @merettm and I have done a lot of work to build that muscle back up." "With all the focus on RL, there's an alpha for us because we think there's so much room left in pre training." "As a result of these efforts, we've been training much stronger models. And that also gives us a lot of confidence carrying into Gemini 3 and other releases coming this end of the year."
Note, "alpha" in the quote above is referring to https://en.wikipedia.org/wiki/Alpha_(finance)
But it's pretty clear that the last full pretrain run they've released is for gpt-4o 2 years ago*, and since then they've just been iterating RL for their models. You don't need any insider information to notice that, it's pretty obvious.
*Excluding GPT-4.5 of course, but even OpenAI probably wants us to forget about that.
> Granite-4.0-Tiny-Base-Preview is a 7B-parameter hybrid mixture-of-experts (MoE) language model featuring a 128k token context window. The architecture leverages Mamba-2, superimposed with a softmax attention for enhanced expressiveness, with no positional encoding for better length generalization. Release Date: May 2nd, 2025
I mean, good for them for shipping, I guess. But seriously, I expect any postgrad student to be able to train a similar model with some rented GPUs. They literally teach MLA to undergrads in the basic LLM class at Stanford [2] so this isn't some exactly some obscure never-heard-of concept.
[1] https://huggingface.co/ibm-granite/granite-4.0-tiny-base-pre...
Also, you're off on Deepseek V3.2's param count, the full model's 685B in size with the MTP layer.
I don't think there's anything interesting here other than "I guess AMD put out a research paper", and it's not cutting edge when Deepseek or even IBM is running laps around them.
[1] Here's a news article from April, although IBM has been doing it for a long time before that https://research.ibm.com/blog/bamba-ssm-transformer-model
I don't know what's so special about this paper.
- They claim to use MLA to reduce KV cache by 90%. Yeah, Deepseek invented that for Deepseek V2 (and also V3 and Deepseek R1 etc)
- They claim to use a hybrid linear attention architecture. So does Deepseek V3.2 and that was weeks ago. Or Granite 4, if you want to go even further back. Or Kimi Linear. Or Qwen3-Next.
- They claimed to save a lot of money not doing a full pre-train run for millions of dollars. Well, so did Deepseek V3.2... Deepseek hasn't done a full $5.6mil full pretraining run since Deepseek V3 in 2024. Deepseek R1 is just a $294k post train on top of the expensive V3 pretrain run. Deepseek V3.2 is just a hybrid linear attention post-train run - i don't know the exact price, but it's probably just a few hundred thousand dollars as well.
Hell, GPT-5, o3, o4-mini, and gpt-4o are all post-trains on top of the same expensive pre-train run for gpt-4o in 2024. That's why they all have the same information cutoff date.
I don't really see anything new or interesting in this paper that isn't already something Deepseek V3.2 has already sort of done (just on a bigger scale). Not exactly the same, but is there anything amazingly new that's not in Deepseek V3.2?
They're just using MLA, which is well known to reduce KV size by 90%. You know, the MLA that's used in... Deepseek V2, Deepseek V3, Deepseek R1, Deepseek V3.1, Deepseek V3.2.
Oh, and they also added some hybrid linear attention stuff to make it faster at long context. You know who else uses hybrid linear attention? Deepseek V3.2.
And hybrid models aren't new, MLA based hybrid models is basically just Deepseek V3.2 in a nutshell. Note that Deepseek V3.2 (and V3.1, R1, and V3... and V2 actually) all use MLA. Deepseek V3.2 is what adds the linear attention stuff.
Actually, since Deepseek V3.1 and Deepseek V3.2 are just post-training on top of the original Deepseek V3 pretrain run, I'd say this paper is basically doing exactly what Deepseek V3.2 did in terms of efficiency.
https://en.wikipedia.org/wiki/Control_theory
This is like arguing if "heater on" or "AC on" is better, which is a pointless argument. That entirely depends on what the temperature is!
Inference is usually less gpu-compute heavy, but much more gpu-vram heavy pound-for-pound compared to training. General rule of thumb is that you need 20x more vram for training a model with X params, than for inference for that same size model. So assuming batch size b, then serving more than 20*b users would tilt vram use on the side of inference.
This isn't really accurate; it's an extremely rough rule of thumb and ignores a lot of stuff. But it's important to point out that inference is quickly adding to costs for all AI companies. Deepseek claims that they used $5.6mil to train Deepseek R1; that's about 10-20 trillion tokens at their current pricing- or 1 million users sending just 100 requests at full context size.
Of course, I doubt you can do that to any random substrate, since the color will depend on the properties of the material.
He's about as "shades of gray" as a politician gets.
Lol. Lmao, even.
No further comment needed.
Well, the devs learnt how to use Github, didn't they? Seems like people CAN learn things that are useful. I can also make the argument that Github pull requests are actually more powerful than git request-pull in addition to having a nicer UI/UX.
Being upset that people aren't using git request-pull is like the creator of Brainfuck being upset that scientists aren't using Brainfuck instead of something more powerful and has a better UI/UX like Python. It's kinda obvious which one is better to use...
To use an example that a regular human would be familiar with: what if the aliens knew math and physics... and were basically ultra-nazis, and very happy to just subjugate you because "our nazi philosophy says that we are superior to everyone else and you are inferior to them" and put you in concentration camps as factory labor for their war machine? You have your own reasons for studying science and math, but what if their entire reason for studying science and math was to build rockets to kill others?
This seems extremely likely, actually! The vast majority of human history has been filled with autocratic governments that centralized power, not free democracies. From the sample size that we have in history, most of the time when the natives meet a stronger scientific power... has not gone well for the natives. What makes you think it will be any different if you meet an alien?
What makes you think just knowing math and physics means that the intelligent alien would be "good" by modern human standards?
And this is just standard boring political talk! We understand that from human politics! What if it's STEM-y and the aliens decide to say "we are killing all of you and slicing all of you into thin slices to scan, in order to scan you into training a LLM"? That sentence would not even be in human thoughtspace 10 years ago! There's almost certainly an even weirder concept that humans today do not have words for, which may be a strong motivation for aliens or even their primary motivation!
I am not an english major, by the way. I am a typical engineer with a strong STEM background, who has happened to have absorbed enough sci-fi concepts through osmosis. I do not consider it likely that we will meet aliens in our lifetime, but I do not expect aliens to follow modern human standards of behavior.
You might as well as argue "we have no idea if aliens exist, being able to imagine aliens does not mean it's actually possible there are aliens", and you'd be technically right... right until the day we meet aliens.
Your line of thought is tantamount to "one should just close your eyes and cover your ears" towards the possibilities in this universe.
Note, I am not a conspiracy theorist and do not believe aliens have visited earth and abducted people or something stupid. But I find it extremely stupid to assume aliens would have familiar moral and ethical systems compared to humans, considering how extremely different human beings already are, and at least humans are all mostly similar! This is similar to european explorers being confused at matriarchal family systems when they meet some random tribe. If some humans cannot even wrap their head around matriarchy, how naive would it be to assume that the average human could be comfortable with alien ethics?
Things off the top of my head that humans usually take for granted as "universals":
- Separation of memory and DNA. What if memories were stored in DNA and can be passed between individuals?
- Inability to share memories. What if memories can be passed around like semen and sweat?
- Inability to easily read others' minds. What if kissing/touching someone would share all of each others' thoughts? How would that alien society develop differently?
- Existence of the ego. What if they live in a constant state of ego death, like some humans on certain drugs?
- Separation of the id and the superego. This is... one way to solve an alignment problem, I suppose. Imagine a species which replaced their sense of hunger/sexual craving, with a craving for morality. And they execute creatures like humans when they see a human do anything immoral, such as eating an ice cream when it can reduce your lifespan and thus deprive your children of a parent, or deprive your society of tax dollars.
- And many other possible examples that i can come up with that exists within human "thoughtspace", let alone concepts that do not exist within human thoughtspace
How would you feel if you met an alien species that communicates by raping their children? If that sounds weird to you, what if they can communicate via the DNA in sperm, so it'd be somewhat similar to how human sex transmits information from the human male to the human female?
Comments may go out of date, but LLMs can quickly generate comments that are up to date. LLMs are more likely to prevent the situation that you described.
> In many cases AIs will even add comments that only make sense as a response to my prompt and don't make any sense in-context.
Yeah, this is a common issue to be fair.
Here's a quote from the paper above:
> Given a list of M models to be served, our goal is to minimize the number of GPU instances N required to meet the SLOs for all models through auto-scaling, thus maximizing resource usage. The strawman strategy, i.e., no auto-scaling at all, reserves at least one dedicated instance for each model, leading to N = O(M)
For example, Qwen2 72b is rarely used these days. And yet it will take up 2 of their H20 gpus (with 96GB VRAM) to serve, at the bare minimum, assuming that they don't quantize the BF16 down to FP8 (and I don't think they would, although other providers probably would). And then there's other older models, like the Qwen 2.5, Qwen 2, Qwen 1.5, and Qwen 1 series models. They all take up VRAM if the endpoint is active!
Alibaba cannot easily just timeout these models from VRAM, even if they only get 1 request per hour.
That's the issue. Their backlog of models take up a large amount of VRAM, and yet get ZERO compute most of the time! You can easily use an easing function to scale up from 2 gpus to 200 gpus, but you cannot ever timeout the last 2 gpus that's serving the model.
If you read the paper linked above, it's actually quite interesting how Alibaba goes and solves this problem.
Meanwhile on the other hand, Deepseek solves the issue by just saying "fuck you, we're serving only our latest model and you can deal with it". They're pretty pragmatic about it at least.