Sysreq2

This is what I was expecting. Very much appreciated. OP’s paper is good - but I sort of feel like it’s singing to the choir. It’s a great resource if you already know the material.

They are most certainly not. It suffers from the nude beach paradox: Anyone who wants you to see them naked, is not someone you want to see naked.

I will say the more recent additions to C++ at least have solved many of my long standing issues with that C-variant. Most of it was stuff that was long overdue. Like string formatting or a thread safe println. But even some of the stuff I didn’t think I would love has been amazing. Modules. Modules bro. Game changer. I’m all in. Honestly C++ is my go to for anything that isn’t just throw away again. Python will always be king of the single use scripts.

Right but until Congress raises the debt ceiling, the Executive Branch has an empty wallet and maxed out credit card. Funding doesn’t go out all at once either. Some programs don’t start until later in the year for example. They typically aim to distribute 1/12th of the yearly allowance per month. Sometimes agencies will over spend early and have to be austere later (like FEMA last year).

Appropriations is Congressional and specifics how much an agency will get over the fiscal year.

Apportionment is under the Executive and addresses when and how those funds are made available.

They are two distinct things.

Considering DOGE is ~100 people and the American labor force is 168 million, in your example, I would be willing to expend 0.000000675% of my time and energy to save 14 cents. That’s the difference between saying keep the change or putting it in my pocket.

Before anyone asks, no I don’t round up for the Boys and Girls Club at Taco Bell. I sometimes feel bad about it though.

I keep harping on this - but two points:

1) Our debt is above GDP and interest is rising faster than GDP growth. Debt is fine when being used as leverage but we are upside down at the moment.

2) We don’t have the ability to issue more debt at the moment. The government has no money. They are using employee pension funds to meet obligations with a promissory note. We have negative cash flow and have run out of the ability to extend our line of credit until Congress raises the debt ceiling.

A lot of people are talking about tool use and writing internal scripts, and yeah, that’s kind of an answer. Really though I think the author is highlighting that LLMs are not being used efficiently at the present moment.

LLMs are great at certain tasks. Databases are better at certain tasks. Calculators too. While we could continually throw more and more compute at the problem, growing layers and injecting more data, wouldn’t it make more sense to just have an LLM call its own back-end calculator agent? When I ask it for obscure information maybe it should just pull from it’s own internal encyclopedia database.

Let LLMs do what they do well, but let’s not forget the decades that brought us here. Even the smartest human still uses a calculator, so why doesn’t an AI? The fact that it writes its own JavaScript is flashy as hell but also completely unnecessary and error prone.

You could also consider using the Common Crawl dataset provided by Amazon. Archive.org is more or less a wrapper around it anyways.

https://registry.opendata.aws/commoncrawl/

SpaceX and by extension Starship are critical for the entire western space industry. Last year SpaceX was responsible for something like 90% of the stuff sent up. Then China and then Russia. There is simply no one else in even in the market right now. SpaceX has lowered costs by an order of magnitude already and hopes Starship will drop it by another.

If you don’t like Elon Musk that’s fine. But his companies get stuff done. Blue Origin has been around longer, has gotten more from the government (Yeah, really) and been getting a billion a year from Bezos with nothing to show for it. It’s insane how much SpaceX has done while everyone else has been doing nothing.

Guys, I’m gonna stop this before it gets out of hand: All we need is love and a shit ton of compute.

Everything else is just details.

So it’s funny you mention that. I actually think Marx made a fantastic critique of capitalism. He kind of nailed the inherent flaws within the system. I compare his solutions to someone predicting that every desk would eventually have a telegraph through. Identifying a problem doesn’t therefor imply a solution.

Regarding the modern Progressive movement though, I find them a bit too authoritarian for my tastes. There are countless examples but classic Western liberalism is at odds with their ideology. They’ve fallen into the classic revolutionary entrapment of their own power being tied with continued upheaval. If you are not with them completely then you are the enemy. From the Google Manifesto to JK Rowling’s cancelling, they remind me of the mob during the French Revolution. They just eat their own and refuse to accept any kind of meaningful critique.

Liberalism requires debate, disagreement and questioning the status quo. I don’t see that amongst the modern Progressives. They would rather burn it down than build something better.

Ironically, I fear them gaining power as much as I do the religious right. Neither contains the capacity for building the coalitions needed for stable governance. I like moderates. Stability. Peace. Compromise. Mutual Respect. Empathy. I just don’t see those values in those circles.

Now that I’ve politely disagreed, I suppose it’s just a matter of time before someone comes along to insult me and prove my point.

I don’t think the modern Democratic Party the same pro-worker, anti-corporate greed version a lot of us grew up with. Don’t take this as an endorsement for Trump mind you. Hardly. Democrats still have basic human rights on their side.

Corporate greed and military spending aren’t up for debate though. We can bicker about guns and immigration policy all we want but some topics aren’t even on the table for discussion.

I don’t think mass adoption is their goal. They had a problem. They solved that problem. They shared how they solved said problem.

Every engineering company releases stuff like this. It’s not meant to change the world. It’s marketing to recruit other engineers who would find that problem interesting.

We’ve been here before: you can’t punish someone into being your consumer. Someone who wants your product at the price you offer it will ultimately pay for it. Short sighted business decisions ultimately hurt the industry more than accepting the need to change your business model.

This reminds me of a conversation I was having recently about genetically engineering humans. I think people are greatly underestimating the complexity of the human body and it’s nature mechanisms of adaptation. Everyone is a little different - but that’s okay because all the little feedback loops and receptor sites account for changes. Like a drug addict needing more and more to get their fix. There isn’t just one variable we can adjust to really solve most problems. You have to look at the entire web of interconnected systems and modulate how each of them work together. We sort of take it all for granted since life is so abundant, but it’s truly nothing short of miraculous.

The odds are getting higher by the day. There is a lot happening in the world with China’s economic blessing. American sanctions don’t work when China manufactures everything.

Relevant: https://samcurry.net/hacking-millions-of-modems

There US Army generally has one tourniquet that it’s deemed best. The CAT by North American rescue. Every Soldier is awash in these things. They hand them out like candy on Halloween. And rightly so.

But the Government has to buy them from businesses that meet certain characteristics. Woman owned, etc. That’s a great thing, right?

In this case however, that female owned business is the guy who invented the things wife. Who he sells them to at a mark-up. Who then sells them to the government.

Economy of scale? Ya! They are the number one buyer and pay the most. Honestly, who cares though, it’s only a couple million a year. Drop in the bucket.

How did Boeing fail? Death by a thousand cuts. That same story probably plays out across the entire supppy chain. Every part, every product, every supplier. Compounded over and over again.

I want to double tap this point. In my experience Claude out performs GPT-4o, Llama 3.1 and Gemma 1.5 significantly.

I have accounts for all three and will generally try to branch out to test them with each new update. Admittedly, I haven’t gotten to Grok yet, but Claude is far and away the best model at the moment. It’s not even close really.

Look, I am not going to install it on my machine and I think a business would be insane to allow it on their servers, but I can also recognize that I am not exactly their market. If I want 10% more performance I can make that happen using my AMEX. I read this and I think “14 year old with a hand me down who wants to be epic” and I can respect that. I remember being a kid and wanting just a little more out of my machine to run the latest and greatest.

While we have started seeing diminishing returns on rote data ingestion, especially with synthetic data leading to collapse, there is plenty of other work being done to suggest that the field will continue to thrive. Moore’s law isn’t going anywhere for at least a decade - so as we get more computing power, faster memory interconnects, and purpose built processors, there is no reason to suspect AI is going to stagnate. Right now the bottleneck is arguably more algorithmic than compute bound anyways. No one will ever need more than 640kb of RAM, right?

And that’s honestly a fair critique. I think Elon is definitely an optimist in his technological ambitions. The whole naming of the drive assistance features from Tesla is kind of convoluted and confusing. I think Autopilot, Fully-Self Driving, all of them could have been termed better. I think he sold a dream before it’s time. And I think the pivot away from using LiDAR was probably a mistake as well.

The maxim of make it work and then make it work better probably applies. He has repeatedly said that since humans only need 2 eyes to drive, AI should too - but that completely ignores the neurological differences between a camera outputting a flat 2d image and the complex make up of our eyes. There is a lot happening there behind the scenes, abstracted away by our brain, that he discounts.

Anyways, yeah, I agree. My only concern is people turning against self-driving because of implementation failures in the short term. Self-driving is going to save lives. Let’s not throw the baby out with the bath water and discount the work being done by people who aren’t Elon. If you seperate FSD from the politics, the business, the fandom, it is an amazing piece of technology. A computer is now driving as well as a person. Like that’s insane. In a different circumstance we would be applauding it in amazement. That team has done amazing work.

You are confusing a claim - specifically that “Teslas are dangerous” or “FSD is dangerous” with a much broader reality; that all Cars are dangerous.

You assert that 3 people died this week due to Tesla related fatalities in the US alone. That adds up to more than 150 per year! But considering that Tesla’s comprise 5% of the car market, that 150 is a drop in the bucket compared to the 43000 that die in car accidents in the US every single year.

Tesla’s fully autonomous self driving is about as safe as a regular person. Many of it’s “accidents” as shown by the various investigations have shown that “drivers” in the plurality of cases had several(>5) seconds to react but failed to do so.

The curious case of Tesla’s safety record will one day be a case study in every undergraduate business class. The naysayers want to see the company fail, the CEO fail, or the technology fail. But the problem is not the technology, but rather the people using it. It’s been long known that increasing protective equipment in sport decreases minor injury but paradoxically increase the likelihood of severe ones.

Why? Because people are lulled into a false sense of security. Individuals take risks they would not otherwise due to the inhibition of the natural feedback mechanism that would otherwise deter them from continuing to push the bounds. Since they keep getting away with it, they continue to push, only to discover that the failure mode has gone from mild to extreme.

Those small injuries are the warnings that keep you safe and prevent you from taking greater and greater risks.

Technology is like that. Cars are safer now than ever. Seat belts, crumple zones, air bags - and yet looking at the statistics you would never know it. This has led to urban legend like cars in the 1940s being “safer” due to their all steel construction. But this is not true.

What is true is that people drove slower in the 1940s. Most highway traversal occurred at a mere 40 miles per hour. A far cry from the 80 everyone does on the way to work today. As safety increases so too does our tolerance for risk.

Ultimately, that is the problem with self driving cars. They work too well. They unintentionally encourage drivers to take greater and greater risks. And they get away with it. Until suddenly. They don’t.

It’s the same story across the DoD. The HUMVEEs are broken down constantly, armor units spend more time trying to keep their tanks running, than actually training.

Someone mentioned that it’s not just Boeing who isn’t the company they used to be. It’s endemic through out American manufacturing. From the latest Boeing failure, Rocketdyne, Chevy and Ford. America doesn’t build anything any more. What it does build is overpriced and not well engineered.

Is it the superstars or the line holders that have been the first adopters? I could speculate, but I am actually curious what you are seeing in practice.

Roko would be proud of you. I welcome our new electric masters.

Exactly what you demonstrated.

‘Meeting Friday’ was the message. You completely ignored the rest. It was just extra padding (intentionally so). Maybe 2 words is too short. But can you honestly tell me that the majority of emails you receive is suscinct and to the point? Or do you simply skim them for highlights and extract what is relevant to you?

That’s really the take away I was trying to get at. People equate quantity to quality far too often. We send way more content than we need to out of fear that someone will equate less with bad.

Soon AI will turn a chickenscrath of notes into a wonderful email. And then turn it back automatically for the end reader.

We put to much emphasis on the look rather than the substance. People are afraid to send out an email with 2 words: Meeting Friday and instead pad it out with pleasantry and detail, context and importance, but none of that really matters.

The nomenclature for microchip manufacturing left reality a couple generations ago. Intel’s 14A process is not a true 14A half-pitch. It’s kind of like how they started naming CPUs off “performance equivalents” instead of using raw clock speed. And this isn’t just Intel. TSMC, Samsung, everyone is doing half-pitch equivalent naming now a days.

This is the industry roadmap from 2022: https://irds.ieee.org/images/files/pdf/2022/2022IRDS_Litho.p... If you look at page 6 there is a nice table that kind of explains it.

Certain feature sizes have hit a point of diminishing returns, so they are finding new ways to increase performance. Each generation is better than the last but we have moved beyond simple shrinkage.

Comparing Intel’s 14A label to TSMCs 16A is meaningless without performance benchmarks. They are both just marketing terms. Like the Intel/AMD CPU wars. You can’t say one is better because the label says it’s faster. There’s so much other stuff to consider.

My last understanding was that it fell outside the standard model and would require some kind of new elementary particle to explain. It doesn’t interact with light or electromagnetism. The only real basis we have for it is gravitational abnormalities.

I guess I have to ask: why do you presume there exists an undetectable force that makes your equations work and not that the equations don’t account for everything?

Seems silly. You would never believe 85% of software bugs were caused by unseen, undetectable stochastic physics errors that only manifested as a system crash. The fact that doing the same thing over and over again produces the same result certainly doesn’t cement that theory in my mind.