> ...one of the truly baffling things from either their team or AMDs team, is how much the workstation grade AMD r9700 has been ignored.
It makes a huge amount of sense after considering AMD's approach to graphics cards from around 2010 to 2025. They just didn't see graphics cards as viable compute platform and many who made the mistake of believing that good specs would translate into in-practice performance got badly burned. I'd have been involved in the AI boom but for an expensive AMD graphics card, I'm not going to forget that for a while.
George Hotz was interesting as a public example, but I think his story probably repeated a few times outside the public eye. People tried to make AMD work and ended up the worse for it.
People who had an interest in using AMD cards to get things done are probably by and large waiting for a new generation of hopefuls to prove this time is different. The mutterings out of AMD are promising, but that isn't persuasive enough given the scale of the failures.
GCN was such a promising compute architecture, AMD even pioneered stuff like async compute and compute shader heavy rendering pipelines, only to never seriously go beyond that on consumer gear.
I agree with your assessment that the story of supporting the competition, only to get burned, has repeated many times with AMD outside the public eye. It's why I don't put much stock in claims that things work great as long as specific flags are used.
You can get those numbers with https://codeberg.org/ggz14/radiance-vllm-mxfp4. I also can get it on a single R9700 but the 75 ~ 80 t/s is only peak acceptance of very predictable tokens like coding or json, and averages lower for prose. It's still much faster than regular llama.cpp.
> Risam’s own concluding words hammer home how data confers more power on the already powerful, over and over again.
I think the insistence of the body politic on giving more power to the powerful is what drives this. Our ability to gather and organise data is just the limiting technical capability. Which is to say, the more data a central authority can process the more power it is capable of wielding.
Unclear why people are so sanguine about that whole issue given the 20th century, but here we are.
Since when has Facebook been associated with MAGA? They were famously part of the attempt to get Trump banned [0] from social media that led to Truth Social. Their moderation team tends anti-MAGA to the point where it is an international political concern.
And, assuming that the article is accurate, obviously they shouldn't have banned an account for posting "Can't handle a pool, but wants our lake?". I think everyone would agree outside the Facebook team. That would be a classic Facebook over-moderation. Extreme to the point where I'm going to assume there is more to the story.
"Oh, being bribed/extorted by the government we took control-of doesn't count, because internally the lügenpresse still has unreasonably-biased employees who keep accusing us of being bad people who'll abuse government power. But don't worry, we will fix that soon..." /s
____
P.S.: I wanted to add something that wasn't ironic, but I'm having a surprisingly hard time unpacking my own sarcastic-intuition into arguments. Perhaps:
1. The MAGA bloc angry at companies moderating Dear Leader have already struck back, and they will not be satisfied by less than a reversal into cheerleading.
2. It can be hard to judge what most employees think when the corporation itself is under some duress.
3. Sometimes the circumstances of a "reversal" demonstrate that the original concern was extremely valid.
The obvious link between the two maps is neither has much concept of landmass size.The Americas are unknown in the first and unsatisfying imbalanced in the later map (Belize & El Salvadore, something must be done!). With all that in mind, it seems possible it could be a coincidence driven by similar constraints on both maps.
The only choice here seems to be whether the containing shape is a circle, a triangle or a square.
>
You can finish reading your friend's favorite book or put it down if you don't like it. Neither are you economically compelled to read it by your workplace or forced to build work on top of it that will become difficult or impossible to modify/maintain if you bin the book.
Economics does sometimes compel you to read books or other material as a condition or expectation of employment. It is difficult to get a good job while being illiterate and many jobs require reading huge volumes of written work. Work in law or journalism springs to mind as economically compelling people to read a lot even if they they might not necessarily want to.
In this context "trust" is something of a euphemism. The US had a big enough military relatively speaking that if they came for your gold then it wasn't feasible to stop them unless you were part of a competing bloc. So, you know. Keep the gold somewhere the US is happy with. Best shot of getting to a good outcome, and the US was pretty reasonable in the 1900s.
That all appears to have changed. They may already have lost the ability to physically coerce even medium powers and at any rate look like they have to retreat the boundaries of their global Not-An-Empire. So it makes sense to start rethinking the gold's physical location.
There is an element of this essay that seems a bit short on the whole rights-and-responsibilities concept that should be highlighted.
On the one hand, she's right that most people would want a wife and she gives a reasonable overview of why. That is why men generally want wives. The wives bring a lot to the table.
If you do the same exercise for husbands, generally most people would also want husbands. The men also bring a lot to the table. Relationships happen when everyone benefits.
Now if she's coming at this from a place of mutual joy and satisfaction then this essay is a pretty good summary of what men are probably looking for in a wife. But I see "instant classic ... [in a] feminist magazine" get the impression that it is probably being interpreted as wives shouldn't have to bring all the things men want to a relationship.
Yeah sure. I can agree with that. They shouldn't have to. Indeed, they don't. As we've discovered in the 1970s-2025 window they can renegotiate the social contract to a large extent. But I'm not really on board with what contribution the essay is supposed to have had beyond highlighting things what men like in women.
Realistically, you meet with one, two, three potential candidates and you realize it's not what you need and want. You could have them, it's just the disadvantages are too high.
> I honestly think most Congress people want to do good for the country.
That seems wildly unlikely. Wouldn't there be evidence for it by now? Over the last decades there have been scant murmurs of the US congress sitting down and taking a serious look at what might be good for a country. I'm happy to say that the root cause is a voting public in the US (everywhere, really) rewarding unprincipled liars and punishing principled politicians. But the outcome is that the US Congress is full of unprincipled liars who have little regard for the good of the country, because that would be a principle that gets in the way of political expediency.
The system isn't some abstract concept foisted upon them from on high. They control it. There are some very high-level constraints on them that are all-weather good ideas and apart from that it is a free-for-all. The reason the system is dysfunctional is because that suits the congresspeople.
Most of the US airforce is executive transport. Clearly you dont actually know people higher up in government. As someone who dated a very high up state legislator in my state said: She totally doesnt care cause she doesnt suffer the consequences. And at the congressional level it is again a level apart. Michal Crichton has a name for this, I think!
> It's just not predicting based on it's training data, but predicting based on RLVR & more, trying to get to the optimal solution ( as much as the solutions CAN be optimal)
It is predicting based on a model. In many cases we can download the model off hugging face. The model is conditioned by all sorts of things. Training data, post-training, coincidence, prompt inputs, runtime data available from whatever means.
> but at least I would still call it a "next token predictor"
We can call any prediction system a next token predictor. If you watch over the shoulder of a human writing a HN comment you are almost certain to see them generating a linear string of tokens. That is what keyboards do. It is impossible to generate text without being equivalent to a next token predictor.
Diffusion LMs denoise a canvas which I personally find more interesting.
I don't really disagree that human cognition is essentially a predictive task though, as I understand it, predictive coding and related theories based on the Bayesian brain hypothesis are fairly popular these days (though maybe not clearly dominant over alterative models? IDK I'm not a neuroscientist). I imagine most people would draft a few tokens before refining them like MTP or diffusion though, if we do decide to use LMs as an analogy to human cognition.
That is logically equivalent to a serial token prediction engine. If you have a diffusion text model you can use it to implement a serial token generator and if you have a serial token generator you can use that to implement a diffusion-generated text string. Don't think about the efficiencies of that, it is an upsetting idea (eg to generate N tokens, the serialising model might recalculate the same string N times from an input and emit one token of it each time - rather wasteful).
It is similar to how everything ends up being Turing complete. Any prediction system has to be equivalent to some sufficiently complicated text generation system to describe the prediction. And any text-generation system has to be equivalent to a sufficiently complicated model that serially emits tokens.
They symptoms sound a lot like humans, so I don't see how it stems from their lack of self reference. Most people you need to keep them in areas they understand or they go to pieces. The lack of self reference just means every time the context clears they reset. They are systems in a permanent state of extreme amnesia.
It’s really not like that. If I ask you to tell me about a geographical place I just made up, you can trivially and generally instantly recognize you don’t recognize it. A child can do this. You wouldn’t be able to hold a job or generally get through life without this level of self awareness.
I had a great game as a kid where me and girls were playing for kisses on "words", i.e. "start the next word with the letter of the word I just said". So I invented a whole new vocabulary and then taught them that vocabulary. Apart from the fact that I got a lot of kisses from them they were so happy they learned something new they immediately went to share this with their (and my) parents. Obviously I got a slap but I just don't get how you can think "a kid can do this". You're either completely ignorant of different cultures (i.e. as an Eqstern European I still find it hard to remember Indian names, let alone tell if they are actually fucking with me) or intentionally simplifying the problem.
No, I’m referring to something else. Not the next word task. Being able to know whether you know something. Next time you encounter a child, ask them something they’re likely to be ignorant of, and see they’re able to say “I don’t know”.
Current frontier models are pretty good at telling you when they don't know something, or if you're asking about something nonexistent or nonsensical. Not always, but there has been massive improvement on this front lately.
If your last experience with frontier LLMs (read: not models that Google and ChatGPT are giving away for free, but models you have to pay for) was over a year ago, you may not realize that.
So can LLMs. I asked one about South Wollopop and it suggested that it'd never heard of it but maybe I meant South Wollo in Ethiopia. And that's just a local model with modest hardware and no internet access. First attempt.
I mean, if our position is that a hyper advanced statistical model is going to ultimately struggle with the concept of something being unlikely to be true then the statisticians may as well give up in despair. There is no theoretical obstacle here.
It makes a huge amount of sense after considering AMD's approach to graphics cards from around 2010 to 2025. They just didn't see graphics cards as viable compute platform and many who made the mistake of believing that good specs would translate into in-practice performance got badly burned. I'd have been involved in the AI boom but for an expensive AMD graphics card, I'm not going to forget that for a while.
George Hotz was interesting as a public example, but I think his story probably repeated a few times outside the public eye. People tried to make AMD work and ended up the worse for it.
People who had an interest in using AMD cards to get things done are probably by and large waiting for a new generation of hopefuls to prove this time is different. The mutterings out of AMD are promising, but that isn't persuasive enough given the scale of the failures.
reply