Symmetry

There's been a lot of churn over the years but additions being done in the same timeframe as XORs has been pretty constant. The Pentium 4 double pumped its ALU but both XORs and ADDs could happen in a half cycle latency. The POWER 6 cut the FO4s of latency in stage from 16 to 10 and kept that parity as well. When you need 2 FO4s for latching between stages and 2 to handle clock jitter at high frequencies the difference between what a XOR needs and what an ADD need start looking smaller, particularly when you include the circuitry to move the data and select the instruction. Maybe if we move to asynchronous circuits?

For a 32 bit number you're looking at going from using 256 to ~1800 transistors in the operation itself. A modern core will have roughly 1,000,000,000 transistors. Some of those are for vector operations that aren't involved in a xor or sub, but most of them are for allowing the core to extract more parallelism from the instruction stream. It's really just a dust mote compared to the power reduction you could get by, e.g., targeting a 10 MHz lower clock rate.

Right. Keeping down the number of slots the scheduler and bypass network need to worry about is an important design pressure.

There's a structure called a carry-bypass adder[1] that lets you add two numbers in O(√n) time for only O(n) gates. That or a similar structure is what modern CPUs use and they allow you two add two numbers in a single clock cycle which is all you care about from a software perspective.

There are also tree adders which add in O(log(n)) time but use O(n^2) gates if you really need the speed, but AFAIK nobody actually does need to.

[1]https://en.wikipedia.org/wiki/Carry-skip_adder

That's all true, but on any modern x86 processor both the single pair of gates for the xor and the 10 or so for a carry-bypass 64 bit wide subtraction both happen with a single clock cycle of latency so from a programmer's perspective they're the same in that sense. There's still an energy difference but its tiny compared to what even the register file and bypass network for the operation use, let along the OoO structures.

There's only a few grams of hydrogen in the reactor's plasma, it's reaction with oxygen wouldn't be much more exciting than just losing containment. There are engineering challenges that have to be addressed but no worse than the 6 MW research reactor I used to walk by every day to my college classes in the middle of a dense city.

The proliferation risk of someone using the neutron flux to produce an atomic or dirty bomb are real but that exists no matter where it is.

"Any problem in computer science can be solved with another level of indirection...except for the problem of too many levels of indirection."

With NVidia/OpenAI actual graphics cards did change hands. Vendor financing, like when a car dealership gives you a loan to buy a new car, is actually pretty normal.

It does have oil. And the reason the UAE is backing the RSF is that they have gold interests there.

On my laptop I have a yin-yang with DRY and YAGNI replacing the dots.

I meant to include both SAE 2 and 3. I think having both lane keeping and cruise control on at the same time will tend to cause people to lose focus in a way they wouldn't if they had to do one or the other.

SAE level 2 is just a bad idea. People can't be expected to carefully monitor a car and take over at a moment's notice when it's doing all the driving. My adaptive cruise control is great and I hope to have a future car where I can zone out while it drives and take over after after a few seconds heads up, but the zone between shouldn't be a valid feature.

If they wanted they could always just double the $/token. They don't seem to be able to keep up with their current demand and that's what companies normally do in that circumstance if they're looking to money grab, no need for the bankshot approach.

"Poison" can also refer to a substance toxic to other animals. We say that chocolate is poisonous to dogs for instance. And a good fraction of Earth's biosphere was killed off by oxygen poisoning in the first of Earth's great mass extinctions.

Also, the dose makes the poison and excess oxygen actually can poison humans. Deep sea divers have to worry about excess oxygen inducing seizures if they mess up their breathing gasses enough. And even 100% oxygen at regular pressure will slowly damage the lungs, something ICUs have to worry about.

Nick Lane had a great book about oxygen, Oxygen, which maybe isn't as good as his book about mitochondria but is well worth reading.

I was misremembering, it's actually cats where hetrochromatic eyes is associated with females. In humans there isn't any difference.

> The technical error that caused accidental chain-of-thought supervision in some prior models (including Mythos Preview) was also present during the training of Claude Opus 4.7, affecting 7.8% of episodes.

>_>

It's more reproducing than surviving. If the population of some species increases and the number of copies of some allele remains constant we could consider than gene less fit than the other alleles, in the population genetics sense. So it's frequency rather than survival that geneticists look at. But that proves that there are indeed other ways that they could have defined fitness if they wanted to.

Likewise humans with heterochromatic eyes are generally women.

Not a tautology but a definition.

https://en.wikipedia.org/wiki/Fitness_(biology)

There actually isn't any locking involved. I can take a new, officially unsupported version of ROCm and just use it with my 7900 XT despite my card not being officially supported and it works. It's just that AMD doesn't feel that they need to invest the resources to run their test suite against my card and bless it as officially supported. And maybe if I was doing something other than running PyTorch I'd run into bugs. But it's just laziness, not malice.

A asteroid has to be absolutely huge to make it all the way down to the surface without slowing down to terminal velocity. Your typical 1kg asteroid will have slowed to terminal velocity dozens of km above the surface. The smaller an object the lower the ratio of its mass to surface area and the more easily it slows down.

I'm guessing running a 200 Hz command rate on an e-series UR which uses 1 kHz internally will give you a protective stop?

It's a good thing in many cases but not if you're going to be running applications distributed as binaries. Maybe if we go the Gentoo route of everybody always recompiling everything for their own system?

It's hard to imagine a student putting together a RVA23 core in a single semester. And you don't really want that in the embedded roles RISC-V has found a lot of success in either.

I think you have a different understanding of what exactly a brand is than do most people. Every profession has the right to develop its own jargon that subtly redefines common words in its own way so I'm not going to say that you're wrong about what "brand" means in any absolute way, just that you're wrong to say that pg is wrong.

Is this the first time internal communications like this have leaked from Anthropic? It'll be unfortunate if Anthropic can't have honest conversations internally going forward for fear of leaks.

That means they're not conscious in the Global Workspace[1] sense but I think it would be going too far to say that that means they're not intelligent.

[1]https://en.wikipedia.org/wiki/Global_workspace_theory

That might be true for the desktop, but RISC-V is wonderful from a pedagogical and research standpoint for academic uses and in the embedded world its license and "only pay for what you need" is also quite nice.

Yeah, the paper launch OP is talking about happened way back in May 2024.

Anthropic's whole worry with mass surveillance was that current law is too loose in the age of AI to offer enough restraint.