Mindbat

The only reason to have an LLM emit sentences like “I understand” is to make it more appealing than a search engine and increase the likelihood that a user will return; that is, it’s another way of maximizing customer engagement.

Sci-fi author Ted Chiang has already written a couple of good essays puncturing the illusions behind the AI industry.

His latest is another masterpiece, using the publication of a “constitution” for Claude by Anthropic as its jumping off point to discuss the absurdity of imagining a program like Claude as a conscious being, and the many ways Anthropic – even as it pushes the fantasy of conscious LLMs on us – doesn’t act in the ways it would have to if such a thing were actually true.

“The conclusion many colleagues are arriving at is that Google’s infra [infrastructure] and eng [engineering] culture was built to be stable and intentionally slow, and that pressures to accelerate using AI are bumping into that.”

Several things from this report from 404 media about Google employees dunking on the AI they’re being forced to use struck me, but this passage, I think illustrates an angle that’s both worrying and sad.

Software companies, if they live long enough, build up a set of internal processes and tools over time. Those processes are often created in reaction to problems: Someone notices Service X goes down in unexpected ways, so they learn to always test system changes under Scenario Y. Someone else realizes that bringing someone from Team C into a project early on means it’s more likely to get a good reception at launch.

This is how muscle memory gets built in a software org. Slowly, painfully, over time. Google is not my favourite of the tech monopolies, but there’s no denying they’ve got decades of muscle memory at this point.

That their current leadership is hell-bent on wiping out that muscle memory in the name of AI is tragic.

Instead of delivering services and software that unlocks value for their client industries, the software industry has spent the past decade or so trying to control their customers and their client industries. Why make software for hotels when you can control the hotel industry? Why make software for taxis when you can replace the entire industry with software?

So many fantastic quotes I could pull from the latest essay by Baldur Bjarnason, but this is my favourite, because it directly articulates something that’s always bothered me about companies like Uber and Airbnb. If they (Uber) really were intending to make taxis better, why didn’t they just write software for taxi companies, and sell it to them? Or write an open-source version, sell customization and customer support, and grow the business that way?

As Bjarnason writes, it’s because those companies were never about the software, they were purely about exploitation and control in the reckless pursuit of growth and returns on investment. A strategy that was only viable so long as US hegemonic power supported it, allowing Uber to ignore local regulations or strong-arm governments into abolishing them.

The whole essay is well worth reading, as it both sketches out how US power underlies the existing tech industry, and how with the rapid decline of that power, we’ve got a chance to build something better.

Fantastic (and brief!) rundown of why the idea of a conscious #LLM is an illusion, and a dangerous one at that:

Rescuers in Lebanon have long grown wary of the double-tap attack, when Israeli forces target a location, wait until people gather to help survivors – and then strike again. Wednesday’s three-wave attack prompted the coining of a fearsome new term: the quadruple tap.

There is nothing in the right of self-defence that can justify the repeated, deliberate killing of rescuers and medical workers. To invent new ways of murdering protected civilians is horrific. #war #lebanon

That’s it up there, fully assembled, Fedora installed, and ready to go :)

It’s the next step in a journey I started last year, when I decided it was time to ditch the Mac and move to open-source.

Switching hasn’t been easy. I’ve been a Mac user for nearly a quarter-century, going all the way back to the days of Bondi Blue iMac and the puck mouse. But the days when Apple was hands-down the best at building hardware and software that “just works” and delights have been over for a few years now. Last year, the steady drop in software quality and simultaneous bad-faith fights with regulators (really, Apple, you couldn’t allow another App Store? or lower the usurious rate you charge devs for your increasingly arbitrary app review?) finally got too much for me. I decided it was time to switch, not to Windows (barf) but to Linux.

So I ditched my iPhone and bought an Android (Samsung Galaxy Pro, it’s…fine. It’s fine.). Sold my Apple Watch and bought a Google Watch (also fine, though I’ve got a Pebble on order I’m excited to try out this spring).

And while I kept my M1 Macbook Pro around in case things got dicey, I bought Lemur Pro from System76 as my first foray into the Linux trenches in a couple decades.

That turned out to be a good short-term fix (the Lemur Pro was relatively cheap), and a way to get my bearings in the modern Linux landscape. But not a long-term solution. The Lemur Pro was under-powered, for one, and constantly running its fans for even the smallest task (two browser tabs? oh no, fans). It’s screen resolution was also a disappointment after living with Macs for so long. But it was light, and had battery life for days, and ran Ubuntu just fine, so it was a nice gateway machine.

A year in, thought, I definitely needed an upgrade. I debated getting a Pangolin from System76, or a StarLabs machine. I tried out two laptops from Tuxedo Computers, even. None of them seemed quite right for me.

Enter the Framework Laptop 13.

Why a Framework?

In brief: longevity.

I’ve been in the habit of upgrading my laptop every year or two. As a Mac user, that always meant listing my old machine for sale, hoping the value had held up, and then dropping down a chunk of cash for the latest and greatest.

Which was both very expensive (that Apple premium) and not always feasible. Sometimes the laptop wouldn’t sell. Sometimes the latest Mac meant signing onto design mistakes (looking at you, Touch Bar) or lemons (cough butterfly keyboard cough) that I wasn’t always willing to splurge on.

But the Framework’s different! Designed from the ground up to be not just repairable, but replaceable. Meaning I can not only swap out the RAM in this thing, or shove a larger hard drive in when I need it, but I can replace the CPU as well. Or the display. Or the keyboard.

In other words, with a Framework I’m not stuck replacing the whole thing every time I need an upgrade. I can go piece by piece, replacing what I want when I want.

Finally, a laptop that can grow with me.

Specs

The rundown: AMD Ryan AI 7 350, 2TB harddrive, 32GB of RAM. Basically, the most I could get and stick with my $3,000 CAD budget.

Assembly

They say it only takes 15 minutes, but that wasn’t my experience. I was careful, the directions are kind of vague, and I was doing it for the first time, so for me it was more like an hour.

And that’s okay! An hour to assemble a laptop isn’t so bad. And oddly enough, because the directions were kind of bad, I felt more accomplishment at the end. Like I’d figured out a puzzle based on a fuzzy photo of the final image. I feel more attached to this machine, not less, because of the mental effort I had to spend in putting it together.

Battery Life

It sucks. Really.

I mean, ten years ago it would have been amazing. But these days, losing 20% charge while in suspend mode overnight is rather lame.

This is especially hard for me, coming from the Lemur Pro, which could run for days on a single charge. Hopefully, Framework will come out with a larger battery for this thing. Until then, I’m keeping the charging cable close.

Update: Okay, the battery life got a lot better when I turned off Bluetooth. Kind of a pain when I want to hook it up to my external monitor and use my wireless keyboard and mouse, but I can kind of forgive it because I prefer typing on its own keyboard anyway (see below).

Keyboard

Simply a joy to type on. For me, the keyboard is better than anything Apple’s produced since the M1 Macbook.

Display

It’s pretty good! I thought the 3:2 ratio would bother me more, but really, I don’t notice it much. Brightness is fine, even at mid-levels to preserve the battery. Text is crisp and clear, which is the most important thing, to me.

Performance

I haven’t run any benchmarks, but so far it seems…fine? I went with the mid-range Ryzen chip, and while it definitely seems better than the chip in the Lemur Pro, it also doesn’t blow me away.

Which is okay! It’s actually more important to me that the fans don’t spin up everytime I open a new webpage, which is definitely what happened with the Lemur Pro on the regular. A nice, quiet, cool-running machine that keeps up with my everyday tasks is, really, all I could ask for.

The OS

I went with the recommended Fedora install. It’s my first time using this distro (well, in anything other than a server, anyway), and so far I’m impressed. It seems a bit more polished than stock Ubuntu, with its workspaces and trackpad gestures. Not so different that I had a large learning curve, but enough of a positive change that I’m probably going to stick with it.

Do I Miss the Mac?

Not really. I miss certain applications, like Fantastical. But for the most part, the software I use either has a Linux version or a capable web app, so I’m able to get everything I need done.

And some of the replacements are better (that is, simpler and more robust) than what’s current on the Mac side of things. I’ll go into more detail in another post, but I think using dnf or flatpaks for software distribution and using Nextcloud for files/tasks/etc is a better experience than Apple’s way.

Conclusion

Our digital lives are actively abusive and hostile, riddled with subtle and overt cons. Our apps are ever-changing, adapting not to our needs or conditions, but to the demands of investors and internal stakeholders that have reduced who we are and what we do to an ever-growing selection of manipulatable metrics…It isn’t that you don’t “get” tech, it’s that the tech you use every day is no longer built for you, and as a result feels a very specific kind of insane.

Fantastic essay that calls out what’s gone wrong in tech in the last ten years, as everyone from Facebook to Spotify pursued “growth” at the expense of actual users.

Congrats to the people of Syria, after a long and terrible civil war: www.theglobeandmail.com/gift/71a5…

It’s my second year to go; last year I left stuffed full of new story ideas and techniques.

Absolutely brilliant essay from sci-fi author Yoon Ha Lee on “authenticity” and the challenge of writing in a culture that constantly wants you to be exotic for them: