The other way to do this, which I literally did n that era, was to just go buy an extra stick of ram out of pocket (Ram used to not be expensive!) and just guerrilla upgrade my work pc myself, IT none the wiser, and just a lot easier and faster than all the BS time and bureaucracy it would have taken otherwise.
I remember even trying this with a cheap HDD->SSD upgrade as well, but by that point they had implemented drive encryption so that foiled that effort.
Literally the best possible kind of psy-op. Same with translating Sesame Street and distributing it to 100s of countries. Same with countries like Thailand exporting Thai food. It's a psy op based on what if we took nice things from our country and subsidized distributing them to the rest of the world, our country would grow in prestige and soft power. And these strategies are usually pretty effective, not to mention usually net-positive for all sides ... compared with so many other horibble harms US and other countries foreign offices have done around the world out of self interest.
All art requires growth and change. Artists can't go on doing the same things forever. Will they take a turn into some dead end...sure. Maybe even for centuries. But remaining static is certain death.
dude take one single art history class... (well, maybe 2 since the first one is going to give you the venus of willendorf and ionic/doric columns...) this is nonsense.
If you need a class to tell you what’s enjoyable to look at and what isn’t, then it’s not a class it’s brainwashing to suppress your natural instincts.
The only reason the CIA did it in secret is that congress _hated_ modern art and killed State Department projects supporting it. The soviets and every other government also promoted their own artists abroad extensively. Not everything the CIA did was a psy-op, sometimes it was a convenient way for the executive branch to do something that congress didn't want to pay for.
I wish those actions made the folks doing that "better" instead of implicating beautiful parts of my childhood in their vicious attempts to keep the entire world under their thumb...
The obvious problem is that when the same group is committing vicious crimes and horrors, it's very hard to look at about anything they do and be like... well, at least folks were getting laid during Midnight Climax.
"Oh, hey, that's cute: the vicious monsters who torture people to death in a global network of black sites are manipulating us again by exporting pleasant videos of puppets teaching children".
what do mean with the example of thailand and thai food, and why is that a psy-op?
Every country tries to subsidize and dump excess production of stuff they produce
> what if we took nice things from our country and subsidized distributing them to the rest of the world
Or the appearance of nice things. Consider the suburb as promoted by American film which has left a very deep mark on the global imagination and that, regrettably, maintains a strong influence over housing markets and urban planning to this day.
This thing has 50% more memory than GTX PRO 6000, which fetch 18k-25k+ each right now on newegg, and generally sold out. If nvdia can still sell out 96GB PCI cards (on a 2 year old architecture) at those prices, how is an AMD equivalent with better specs dead on arrival?
Market takes time to catch up. When released Qwen 3.6 had a lot of sense and it did fit, but frontier since went too far ahead. Now you'd want DS V4 Flash from yesterday, or you are too far behind.
Unless MI350P will be much cheaper than 6000 Pro, which is unlikely given its VRAM, it will lose to it, because you'd need 2 of either.
Don't see how it could be cheaper with HBM than the RTX 6000 Blackwell Server Pro but in these times of relative shortage any additionnal supply should be slurped ?
I was remarking on the MI350P because I've had a hard time procuring "small" CDNAx systems (for e.g. development, experiments and lower-profile servers) and OAM seemed very niche (not if you're aiming for density and training/inference...).
I hope the MI350P fills a lower part of the spectrum and I can start massively porting CUDA stuff or at least work on HIP and ROCm and what I need to make most or some of our CUDA stuff run on AMD HW, then how make it run fast.
Okay this tidbit is interesting to me "5–6 tok/s M2 -> 31–35 tok/s on an M5 Pro". So where will be in just another gen or two?
my impression right now is that M5 gen is on the cusp of practicality for local inference.
If techniques like OPs here, start to make the RAM situation more amenable, by the time we get to M6 or M7 (or AMD's equiv next gen APUs on TSMC N2 nodes), local AI could be ready to go much more mainstream.
Basically what we're looking at by the M7 generation is a tier shift, where the base M7 can do what the M2 pro did, and every tier moves up accordingly, with the M7 ultra becoming competitive with nvidia dedicated consumer hardware.
My assumption is that the difference is 90% from more memory. I'm making several assumptions because nothing here looks groundbreaking so I don't care to dig deeper, but the model + KV cache definitely cannot fit in memory on the 8GB machine, but probably can on the 24GB machine—or can at least get close. Assuming that this benchmark makes use of that, skipping SSD streaming will speed things up massively (I would have guessed much higher than the reported 6x speedup).
I have an M5 128GB. Being on the cusp of practical is a good description. It will run, but prefill and token gen are still slow relative to my consumer GPU box.
It also gets very hot. If you’ve never heard the fans on Apple Silicon really spin up, it could surprise you. Makes the full GPU setup feel quiet by comparison.
I think after the hardware market calms down the ticket is going to be a light laptop with a second dedicated inference server on the network.
Well that's a relief, glad I am a turtle then and so are most of my friends, and that our whole world economy and food system runs on turtles. So we should be fine.
There’s always a buyer on the other side of a short, naked or located. The question is whether the originating broker actually borrowed the shares.
They are supposed to verify that before they place the trade and generally do follow the rules. Because if they don’t, they will be not allowed to allow any short sales for that security.
> They are supposed to verify that before they place the trade and generally do follow the rules. Because if they don’t, they will be not allowed to allow any short sales for that security.
SEC is clearly ignoring FTD’s and as a result allowing MM’s to reloan unlocated synthetic shares.
Which is exactly what happened in 2008 with mortgage backed securities and CDO’s.
They're illegal, and you can look at them just as the broker making a long bet themselves. Since they'll have to pay off the short seller, if the stock goes down.
"Latent space representation" I have been waiting for this moment in the evolution of AI. Well, waiting with some trepidation. It seems inevitable that frontier AI's will, at some point, leave behind human-comprehensible representations of language. Purely for functional reasons, it's going to start making sense for AI agents to communicate amongst themselves in much more efficient ways than borrowing the languages of flesh-bag humans as an interface medium.
I Imagine next that programming languages, interfaces and API design starts going this direction next. Being written, expressed and optimized as blobs of high dimensional vector space. As humans we might still be able to understand some abstractions of what our AI's are talking about to each other, but maybe not more so then we understand how different regions of our own brain communicate with each other.
I strongly believe that the future is the other way. New programming languages and environments designed for strong auditability and preventing bugs will dominate. Only bad actors will use latent space representation, and it might even be outlawed. But the bad actors will proliferate underground…
Even for like token efficiency it could make sense - like imagine if the representation were more compact
Agents acn already translate languages quite well. It doesn't seem crazy that they could work and think in a model specific language, and then translate back to English or something for the user
It seems like the most efficient method would be for LLMs to communicated by exchanged latent space representations directly. Serial language is a incredibly inefficient way to encode these, a lot like flattening a complex graph into text.
this strategy worked to keep Ukraine alive, by enabling them to throw literally anything and everything they could obtain into the fight. And the system enabled rapid experimentation and evolution of what works. Also they didn't have enough of anything to equip all units equally or fully, so a market-like system of was also a way to triage short supply.
However the logistics costs of fragmentation are very real (relevant to the supply chain theme of this story). And now that Ukraine is producing the better part of 10 millions(!) of drones per year, they are shifting towards more standardized drone models to simplify logistics, achieve more economies of scale and also now to have the capacity to keep units equipped more evenly and reliably.
Reminds me of the Cambrian revolution: suddenly there were all kinds of weird animals. Many of these kinds rapidly disappeared, while a few more successful ones kept on. Or at least that's my reading.
Look at 1950s aircraft. That was the decade of really weird aircraft, as people figured out how jets were supposed to work. Supersonics. VTOL aircraft. The X-planes. Rocket-assisted takeoff. After that, more was known about what worked, and designs became more similar.
There's something hilarious, in the truly cosmic sense of the word, about discussing an "explosion" that spanned between 10 to 25 millions of years of its duration.
Wonder how xenoanthropologists will discuss the "simian explosion" that we're currently experiencing (barely 300,000 years old ATM).
Wouldn't a fragmented, decentralized system also help make their supply chains more resilient? If they had a single large drone factory, it would be a sizable target.
During WW2 in the United States, you had all sorts of consumer goods companies reorganized to output a prodigious amount of military supplies. There were multiple companies making the same model of things, with fairly rigorous QA to ensure quality and uniformity.
For example, the BC-348 receiver, widely used in aircraft, was produced initially by RCA, and eventually "farmed out" to 3 other manufacturers.
More than 4 million M1903 Springfield Rifle were produced by the Smith-Corona typewriter company.
Here's a really good example, look at how the production of proximity fuzes, was distributed.[1]
The key thing is to have second sources for everything. Something the US military seems to have forgotten, or decided to ignore in their pursuit of gold-plated weapons systems that give the most kick-backs.
It's not a great comparison because Germany could not hit the US mainland. Even if there had been a single giant everything factory it wouldn't have mattered.
IIRC the US did plan in case either Japan or Germany somehow became capable of bombing the states. Some factories had their roofs painted to make them blend in with the surroundings, others were built out of reinforced concrete to make them bomb proof
reply