arcanus

Conversely: in humans, intelligence is inversely correlated with crime.

It doesn't go to zero, however!

Hopper had 60 TF FP64, Blackwell has 45 TF, and Rubin has 33 TF.

It is pretty clear that Nvidia is sunsetting FP64 support, and they are selling a story that no serious computational scientist I know believes, namely that you can use low precision operations to emulate higher precision.

See for example, https://www.theregister.com/2026/01/18/nvidia_fp64_emulation...

It seems the emulation approach is slower, has more errors, and doesn't apply to FP64 vector, only matrix operations.

How can someone say "the world is in peril" and not be sounding the alarm? Shouldn't they be calling for action across society?!?

That's a surprisingly large overhead. I've not measured that large an impact on AMD, particularly for compute heavy.

Did you profile at all? And have you observed if it's not compute-bound? If it's memory or IO bound it can be due to other virtualization overheads, such as memory encryption.

The manufacturing cost is emphatically not only 1% of the total development cost. Particularly for GPUs, the high bandwidth memory and manufacturing costs are a significant portion of the product price.

NVIDIA's control over the GPU compute ecosystem — particularly in AI, HPC

The two largest supercomputers in the world are powered by AMD. I don't think it's accurate to say Nvidia has monopoly on HPC

Source: https://top500.org/lists/top500/2025/06/

I do not find her critique of argument #2 compelling [1]. Monetization of AI is key to economic growth. She's focused on the democratic aspects of AI, which frankly aren't pertinent. The real "race" in AI is between economic and financial forces, with huge infrastructure investments requiring a massive return on investment to justify the expense. From this perspective, increasing the customer base and revenue of the company is the objective. Without this success, investment in AI will drop, and with it, company valuations.

The essay attempted to mitigate this by noting OAI is nominally a non-profit. But it's clear the actions of the leadership are firmly aligned with traditional capitalism. That's perhaps the only interesting subtly of the issue, but the essay missed this entirely. The omission could not have been intentional, because it provides a complete motivation for item #2.

[1] #2 is 'The US is a democracy and China isn’t, so anything that helps the US “win” the AI “race” is good for democracy.'

And when I reach the part where the AI, having copied itself all over the Internet and built robot factories, then invents and releases self-replicating nanotechnology that gobbles the surface of the earth in hours or days, a large part of me still screams out that there must be practical bottlenecks that haven’t been entirely accounted for here.

This is the crux of the issue. There's simply no clearly articulated doom scenarios that don't involve massive leaps in capabilities that are explained away by the 'singularity' being essentially magic. The entire approach is a doomed version of deus ex machina.

It also appears quite telling the traditional approach is focused on exotic technologies, such as nanotech, and not ICBMs. That's also magical thinking.

The overall funding for starting up a laboratory is much smaller in Europe than the USA.

It's that start up funding that will drive talent, more than anything else.

While FP64 is indeed important for supercomputers, the largest supercomputers have a great deal in common with AI infrastructure.

For example, high bandwidth, low latency interconnects, supporting GPU direct network messaging and IO, are important.

High memory bandwidth is also quite important.

Debugging and performance profiling at scale also commonly uses similar tools.

As these are related to national defense and nuclear security, I've heard they are largely safe.

Thanks! That's consistent with my impression.

Amazon gets a lot of flak for totally bungling their internal AI model development, squandering massive amounts of internal compute resources on models that ultimately are not competitive, but the custom silicon is another matter

Juicy. Anyone have a link or context to this? I'd not heard of this reception to NOVA and related.

"You may think you know how much Intel is struggling, but the reality is worse."

this would be a perfect use case for the world’s largest, most powerful government which flaunts his values of freedom, democracy, and other American values to inject its vast resources

https://www.energy.gov/fasst

Frontiers in Artificial Intelligence for Science, Security and Technology (FASST)

DOE is seeking public input to inform how DOE can leverage its existing assets at its 17 national labs and partner with external organizations to support building a national AI capability.

To first order, if doctors work fewer hours, more doctors will be required to serve the aggregate medical demand.

More doctors working on the same amount of medical demand will be an increased supply of doctors, and so will reduce doctors salaries.

This seems like an ideal outcome for consumers, and I suspect the AMA would lobby against this. But American doctors should also be aware that their famously world class compensation might be reduced with this tradeoff.

I don’t want an overworked, stressed doctor. I want a peaceful, calm, thoughtful doctor with a happy home life.

I agree but doctors should go in with open eyes that working fewer hours will require more doctors, and this will in turn drive down MD salaries.

I think that's a great idea, but that seems to be the undeniable tradeoff.

The study appears to be misrepresented. From the paper at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10274991/

"Although biological ageing was accelerated in sedentary and highly active classes, after adjusting for other lifestyle-related factors, the associations mainly attenuated."

"Conclusion: Being active may reflect a healthy phenotype instead of causally reducing mortality."

Therefore, the paper isn't supporting the article claim that too much exercise is speeding up aging, rather, that this is dominated by genetics.

I'm skeptical those numbers work out. Winning any amount of $ in a lottery is vanishingly small odds.

Even if true, that would imply ANY probability of a comfortable lifestyle is ~0.

Perhaps the claim is that Moore's law hasn't ended, but it's sweating!

Who needs GPGPU anyway, right?

The world's largest computer, Frontier at oak ridge national laboratory, runs AMD GPUs. AMD is undisputably the #2 in the GPU space.

https://www.top500.org/lists/top500/2023/11/

Some of the world's largest super-computers / HPC applications do use SYCL for AMD! The application I'm most aware of for this is GROMACS. As to why? - because having 3 version of the same code using different programming APIs is a big maintenance burden.

The fact that GROMACS is unwilling to drop CUDA support to stand fully behind SYCL is very telling.

SYCL is a terrible option. The performance of any application on SYCL is currently quite poor. Why drop ROCm (used on the world's largest supercomputer?) for an even more unproven product that AMD has no control over the roadmap?

Banding together with Intel to support SYCL would in my opinion

Except that Intel has more control over SYCL and has repeatedly hurt AMD products with anticompetitive behavior in the past. Why would AMD permit their software to be controlled be a competitor?

AMD is executing with ROCm NOW, with multiple Top500 supercomputer wins and deployment in major cloud vendors with MI300X. Yes, the software needs to improve, but the practice of throwing out software to start over is not a good strategy.

AMD has far more momentum in the data center than INTC at present.

Soon. The card is coming in Q4. The early shipments are likely all going to LLNL's El Capitan Exascale computer: https://www.tomshardware.com/news/amds-instinct-mi300-moves-...

AMD has the hardware but the support for HPC is non-existent outside of the joke that is bliss and AOCL.

You are probably two years behind the state of the art. The world's largest supercomputer, OLCF's Frontier, runs AMD CPUs and GPUs. It's emphatically using ROCm, not just BLIS and AOCL. See for example: https://docs.olcf.ornl.gov/systems/frontier_user_guide.html

That's hardly non-existent support for HPC.

HIP is an API equivalent to CUDA. The kernel code is identical. The host code is the same API with a change from the CUDA to HIP namespace. This seems to be an extremely minimal form of 'porting'.

But can it run PyTorch?

In office jobs, by contrast, productivity remains rooted in notions of busyness and multi-faceted activity. The most productive knowledge workers are those who stay on top of their inboxes and somehow juggle the dozens of obligations, from the small tasks to major projects, hurled in their direction every week.

I disagree. I think this looks productive, but people who only do this will hit a terminal level and find themselves stuck.

Getting above this, into Principle/Director/Executive ranks certainly requires being a great communicator, but also typically involves __success__. I see this as shipping a successful product, closing a key customer deal, etc.

It's easy to see productive people and expect that being on top of email and product plans is what made them successful, but this is only an effect of their productivity and mastery of the product development.

Note that John Wick did have great production values: it needed much more than 'just' Keano. That logistics happened in the background.

I'm against gatekeeping. Visualize success. Be who you want to be.

The sooner I called myself a scientist in life, the better.

L9s ship more ppt than code