cowmix

Side note: I tried the same questions with some local LLMs I’m running at home—unfortunately, they’re nowhere near as good or useful. I hope local models improve quickly, so we’re not left depending on the good graces of big LLM(tm).

This article feels “ripped from today’s headlines” for me, as my mother-in-law was recently in the ICU after a fall that caused head trauma. The level of AI-driven automated decision-making is unsettling, especially as it seems to allow large organizations to deflect accountability—“See? The AI made us do it!” I’m not entirely sure what guided her care—or lack thereof—but, as someone who frequently works in healthcare IT, I see these issues raised all the time.

On the other hand, having access to my own “AI” was incredibly helpful during her incident. While in the ICU, speaking with her doctors, I used ChatGPT and Claude to become a better advocate for her by asking more informed questions. I could even take pictures of the monitors tracking her vitals, and ChatGPT helped me interpret the readings, which was surprisingly useful.

In this “AI-first” world we’re heading into, individuals need their own tools to navigate the asymmetric power dynamic with large organizations. I wonder how long it will be until these public AI models get “tweaked” to limit their effectiveness in helping us question “the man.”

"I hope this message finds you well." --- busted!

The best family trip my wife and I did with my kids (8 and 12) was Amtrak from AZ -> DC -> NYC. It was relaxing, fun and different. We even did it in coach (so no private room) and it was still amazing.

We flew home -- and even though the flight time was about 5.5 hours - it was stressful and a let down. Your family's milage may vary.

I don’t think ultra-wealthy preppers want catastrophic events to happen, and I agree that prepping in itself isn’t nefarious. My concern is more about the mindset it can encourage. When someone like Altman—who has significant influence over the future of technology—starts focusing on a solid “Plan B,” it might lead them to take more risks, consciously or unconsciously. Having what they believe is a safe fallback could make them more comfortable pushing boundaries in ways that could accelerate instability. It’s not about wanting disaster, but rather how preparing for one might subtly shift decision-making. For instance, “AI safety team.. who needs it? amirite?!”

I agree that prepping itself is completely fine, and people should be prepared for natural disasters, civil unrest, or whatever scenarios they’re most comfortable with. Building resilience is beneficial for both individuals and communities, and I can see how it plays an important role, especially in situations like the ones you mentioned in NC.

My issue, though, is with someone like Sam Altman—a leader of an organization that could potentially accelerate the downfall of civilization—being so deeply invested in prepping. Altman isn’t just a regular guy preparing for emergencies; he’s an incredibly wealthy individual who has openly discussed stockpiling machine guns and setting up private land he can retreat to at a moment’s notice. It’s that level of preparation, combined with his position at the helm of one of the most consequential tech companies, that doesn’t sit well with me. It feels like he’s hedging against the very future his company might be shaping.

I keep thinking about Sam Altman’s March ’23 interview on Lex Fridman’s podcast—this was after GPT-4’s release and before he was ousted as CEO. Two things he said really stuck with me:

First, he mentioned wishing he was more into AI. While I appreciate the honesty, it was pretty off-putting. Here’s the CEO of a company building arguably the most consequential technology of our time, and he’s expressing apathy? That bugs me. Sure, having a dispassionate leader might have its advantages, but overall, his lack of enthusiasm left a bad taste in my mouth. Why IS he the CEO then?

Second, he talked about going on a “world tour” to meet ChatGPT users and get their feedback. He actually mentioned meeting them in pubs, etc. That just sounded like complete BS. It felt like politician-level insincerity—I highly doubt he’s spoken with any end-users in a meaningful way.

And one more thing: Altman being a well-known ‘prepper’ doesn’t sit well with me. No offense to preppers, but it gives me the impression he’s not entirely invested in civilization’s long-term prospects. Fine for a private citizen, but not exactly reassuring for the guy leading an organization that could accelerate its collapse.

My plan was to run the laptop Windows and the 'devkit' Linux. This first batch of laptops (AFAICT) can NOT dual boot.

My thinking is this, if Windows ARM is a success -- there will be more units out there that can ALSO run Linux too. If Windows ARM is a failure, then Linux will suffer too.

A large part of the success of this new platform is how fast devs can adapt / fix their apps to work natively. Apple, for instance, provided dev mules for OSX ARM --- and their rollout of Apple Silicon was smoother than anyone could have hoped.

Windows ARM -- still borked in SOOO many ways -- and its 10+ years old now.

When this platform was heralded as the “AI” desktop, I pre-ordered both the dev kit and a laptop. Like many of you, I’ve experienced a months-long delay in the delivery of the dev kits. Although I STILL don't have my devkit, I received my laptop pretty much on time. -- and I quickly discovered that despite Windows on ARM (WOA) being over a decade old, the support for open-source tooling is as complete as Swiss cheese. Key Python modules are missing, and even the Git command-line (git bash) client isn’t functional yet!

I mean, forget about basic open-source development, let alone performing AI inference work on your new Snapdragon laptop.

After some digging, I’ve learned that just four overworked developers in Prague make up the core team unclogging this tooling dependency log-jam. Gah!

For what it’s worth, WSL2 (Linux on Windows) actually runs quite impressively on the Snapdragon X.

When I had my ISP back in the early 90s, everyone in the office registered all those "generic" domains (along with trademarked ones too). My ex-business partner originally registered beer.com, his sister did game.com (https://www.youtube.com/watch?v=PTuQgUe-CME), our billing manager did subway.com and someone even did wallstreet.com.

Those were the days.

Pytorch alreadh has Vulkan support -- and Kompute does not support pytorch yet. That's is going to show adaptation of this project.

I rushed out and bought one of the new Snapdragon laptops – the Lenovo Yoga Slim with 32GB RAM. I figured that since Microsoft marketed this as an "AI" laptop, it would be a perfect platform for ML development. However, my experience has been the complete opposite. Python for Windows ARM64 is a complete mess. You have to build almost every module from scratch, and it’s a nightmare of broken and buggy dependencies.

After 10+ years of Windows ARM, I find these basic toolchain gaps inexcusable.

For what it’s worth, WSL on Windows ARM64 is pretty much great.

Arizona State University was an early (first?) adopter in providing broad access to ChatGPT Enterprise, but the implementation has not been as smooth as I hoped. My wife works at ASU, and from our discussions, it’s clear that the rollout has been somewhat disorganized.

The promotional announcements led me to expect "ChatGPT 4 for all!" Yet, the distribution of access has been ambiguous, leaving many staff members, including my wife, in the dark. There has been little to no guidance on how to effectively integrate ChatGTP into their daily tasks, appropriate use cases, or any potential costs involved.

The gap between the initial excitement generated by press releases and the reality of the rollout's execution has been significant. Even now, the specifics remain unclear, despite my efforts to assist my wife in navigating this whole thing.

I was always under the impression that WSL was good on RAM - giving it back when not needed. There was never a way to tune RAM usage AFAIK.

A lot of my right of center friends have seemed to justify drug use just fine these days.

When my wife and I do our morning walks around downtown Phoenix in the early AM, we're always amazed at how much weed we can smell. I mean, 6:30am -- a good part of the city smells like a Bob Marley concert -- how do people get anything done these days?

VideoScape 3D and Sculpt 4D where my bread and butter at the time. Although anything would crash.

One day I'll rip and upload a video I made in the late 80s of when I was typing up a chemistry lab report. My Amiga was producing so many guru meditation errors I had my VCR record my typing (in real-time) so I had proof that I did the report. I handed in a draft of the report and a VHS tape.

As someone who followed the development of the Amiga and then bought the Amiga 1000 (plus everything else I could afford with my paper route earnings) the first day it was available -- I do NOT yearn to relive those days. I loved it then and as documented everywhere, it was a head of its time. However, the instability of software and how much Commodore didn't iterate on the technology -- that's what killed the Amiga. I still have PTSD from all the insane amount of crashing and rebooting my setup did.

500+ people to even do the HIGH level bits of running a SC network isn't a lot.

Everyone (even the most strident Tesla haters) agrees that SC is the one thing Tesla does the best, hands down. I own a model Y and tried to use only non-Tesla charging on a long distance trip a few months ago, it was a disaster.

Telsa's charging network is a win on EVERY front:

1. Locations 2. Quantity of locations 3. Quality (high charging rates) 4. User experience / design of hardware - software 5. Realtime reporting and navigation 6. Uptime of network

The SC network is why a lot people consider Tesla who otherwise it would be a big fat no.

Holy crap.. I stand corrected.

Gah! That headline sucks. It makes is sound like he laid off EVERYONE that works on those parts of the company. No, he sacked the head of those efforts, right?

Musk is still nutty and erratic, but lets get the headline correct people.

This still has the best SOCKS5 proxy abilities, even over normal Firefox. This browser + "ssh -D" has saved me many times working in highly restrictive networks.

When I had my little ISP here in Arizona back in the early 90s, I gave every new customer a copy of the Cuckoo's Egg.

We all die alone. God is dead. Etc.

Is this a comment on how one must use Javascript to program the JWST?

Yup, the quality in how these programs get upscaled for HD is all over the map. There's a lot of examples of people on Youtube and archive.org who are ripping their personal collections of old content who are doing this, I would say, correctly.

Here's one of my favorite examples of archiving being done "right":

https://archive.org/details/@the_vista_group

t-shirt ordered.. thx!

I don't like the subscription options they offer. I just want to throw them $50 a year for the CE version -- which I would think is better than $0 a year I'm giving them now.

DBeaver is amazing. As someone who needs to do adhoc querying / extracting / loading of data from any hosts of system on a daily basis - this tool has saved me over and over again.

My beef is there doesn't seem to be a way to contribute $$$ to the OS version -- except for buying/subscribing to the commercial version. Maybe I've missed some web page that explains how to do contribute -- so if anyone knows if there's a way to do that --- post it here.