Most of the country is genuinely committed to the bill of rights. The Trump administration is determined to ignore every single amendment, but even a lot of the Republican party I don't really think wants this. People are genuinely worried about Chinese media control. But Trump obviously wants to control the media and censor things. I hope the right turns around. Assuming that everyone in politics is working in bad faith is how we become an authoritarian country like China. It is hard when the leadership is obviously working in bad faith and the entire Republican party deliberately chooses bad faith and lies over any reasonable alternatives.

> Most of the country is genuinely committed to the bill of rights.

I'd like to see evidence of that. A third of the country voted to burn the bill of rights, and another third voted they don't care but they'd be ok with it happening.

US government policy is completely aligned with the goal of stopping Iran from doing this, there is no reason to protest the US government on this issue.

It's not always a protest against government, sometimes it a campaign of lobbying, sometimes it's international attention.

The US government wasn't a friend of Kony in 2012. Before Trump 2, the US were not that friendly with Russia, yet people protested in many places around the world to show support for Ukraine and to voice their opposition to Russia's imperialistic wars, being aligned with their governments' position.

It's different with Iran. Some of that is likely to be Iran's lower profile, but not all -- it's not like media outlets are not reporting on it at all and you have to get your information from niche sources to hear about events in Iran.

I had a phone I liked that was stuck on Android 7 and had increasing sync issues until I got a phone that can run current Android. iPhone should be better with support, but also, Apple is hostile to third-party apps that use Bluetooth, so I'm hesitant to say this is Garmin's fault and not Apple's.

They actually note the opposite of that, which is that fewer elderly people are living alone.

> I get that, but I do witness a lot of compassion and help directed to homeless folks. However, even if they're regularly gifted by strangers, it's likely not enough to materially change their situation.

When I've looked at the data, the majority of homeless people have been homeless less than 12 months. This means that the majority of homeless people who benefit from support will use it to get out of that situation quickly. And for the most part, if you give help it will be immediately and materially useful.

I wouldn't say exponential, I would also say it's likely that Amazon is projecting demand to rise faster than it is. (But it is still consistently rising, and with demand rising they would prefer to overestimate than underestimate.)

The thing historically about GPUs has not been the actual lifespan of the hardware (at least half of the hardware will probably work fine for 10 or more years) the problem is that work/watt is dropping for newer hardware, so there's a point where even if you had an equivalent quantity of 10-year-old GPUs, powering them for some period costs $40k and you can buy a single brand-new GPU that costs $40k but only costs $20K to power for the same period which is less than a few years.

I don't think we're seeing any decrease in supply though, ignoring 2020 I'm pretty sure the number of GPUs manufactured has been steadily increasing. It might be the case that projected manufacturing was higher than what actually happened, which is not the same thing as a decrease in supply, but companies like Amazon will talk about it like it is, and from the standpoint of their pricing it essentially is.

> the problem is that work/watt is dropping for newer hardware, so there's a point where even if you had an equivalent quantity of 10-year-old GPUs, powering them for some period costs $40k

Sell the old-gen GPU's to on-prem users (including home consumers) who are going to run them a small % of the time (so power use is more or less negligible to them compared to acquisition cost), problem solved.

The same math applies for on-prem/home users. If you actually have some workload where it makes sense to get a free GPU that costs $40/hour to power because you only need it for a few hours a month, it's probably cheaper to rent a more efficient GPU from someone who can power it at a lower cost.

I would not be surprised if the majority of H200s were manufactured in the past 12 or even 6 months.

I've been prototyping using LLMs for some borderline use cases, and the cost isn't really the concern, it's the reliability. Using less than the most frontier model seems irresponsible if it could mean the difference between 99.95% reliability and 99% reliability, and that's the threshold where you should've hired a human to do it because you lost more money on that 0.95% error rate than you saved on salaries. (I don't actually have any use cases where this kind of calculation makes sense, but in principle I think it applies to most uses of LLMs, even if you can't quantify the harm.)

Problem is that the frontier models are nowhere near 99% reliable. Orchestration and good system design is how you get reliability. Yes, the frontier models still are going to be better by default than open source models. But the LLM is still only a component in a broader system. What's seeming to be actually necessary for any high-usage worthwhile use case is making your model task specific (via fine-tuning / post-training / RL). I build these systems for enterprises. The frontier models are not enough.

Is that really true, though?

First off, you’re ignoring error bars. On average, frontier models might be 99.95% accurate. But for many work streams, there are surely tail cases where a series of questions only produce 99% accuracy (or even less), even in the frontier model case.

The challenge that businesses face is how to integrate these fallible models into reliable and repeatable business processes. That doesn’t sound so different than software engineering of yesteryear.

I suspect that as AI hype continues to level-off, business leaders will come to their senses and realize that it’s more marginally productive to spend on integration practices than squeaking out minor gains on frontier models.

Using a tool like notion is organizing your thoughts into a coherent structure so that you can reexamine them. Running that coherent structure through an LLM replaces your structured thoughts with other thoughts you didn't have. You've gone to all this trouble to record your memories so they don't change, and then you run them through a machine that replaces your memories with randomly generated ones.

I think AI is a great tool in certain circumstances, but this sounds like one of the clearest examples where it is brain rot.