Ryan Egesdahl

Picking a fight with an astronaut is only slightly safer than picking a fight with a librarian.

I was in at around the same time, and we got maybe twice that, with a brief mention that it would be on us to show the order was unlawful. That doesn’t exactly encourage enlistees to refuse an unlawful order while being pressured by their entire chain of command to follow it.

So ok I joined the Navy in 1998 and maybe things have changed since then but I got approximately 5 minutes of training on what constitutes an illegal order and as someone involved in a weapons system *I did not have access to intel on who I was firing on until afterwards*.

It’s *almost* like their understanding of the world comes from their rage over the world not revolving around them.

I don’t always (read: often do not) agree with you, but at least I’ll actually read what you have to say before responding to it. Geez, some people… 🙄

Just a friendly reminder as we go into the holidays:

Folks don't have to explain to you why they don't drink.

If you offer them a drink and they decline, don't ask them why. It's not your business.

Just offer them something without alcohol to drink instead. And move on.

I get a strong impression that efficiency wasn’t the real concern behind their response…! 😮

I’m an atheist with a similar - if not identical - viewpoint, so I’d say it’s a reasonable way to approach human complexity. ❤️

It’s *so* tiresome, too. 😮‍💨

The problem with Christian fundamentalism is the beliefs in univocality and inerrancy of scripture, which leads them to some truly bizarre conclusions about the texts. Most of the time, they just end up excusing their own hateful beliefs with what the texts supposedly endorse.

Yes, I agree with this. The texts are contradictory, both within and between each other. People reading them must negotiate with them and decide which parts to emphasize or minimize to arrive at a set of beliefs around them. There’s no way to believe *all* of it at once.

Yes, that’s what the canon plainly shows. It’s fundamentalism when people believe that whatever God does is definitionally good, though. That seems to be what Dale Partridge promotes, but most of Christendom would disagree with his viewpoint for one reason or another.

I think there’s a lot of confusion among people in the US as to what “most of Christendom” looks like, and vice/versa for those elsewhere. Christian fundamentalism is popular in the US (and parts of Canada), but it’s looked upon with bemusement elsewhere.

Like… you print things in batches…? So why not just print batches in different colors? You’re going to “waste” some filament every print cycle anyway, so why not just take advantage of that and do something fun with it? There is no significant increase in efficiency with less fun here.

I’m not a SME by any means, but I am aware of computing history, and in particular the development of AI throughout it. Based on what I know, yes, I think expert systems are the way to go. We learned a long time ago that computers cannot reason like humans, and we’ve proven why that is as well.

Truly evergreen.

Fair enough, I suppose. It’s just… well, it’s obvious that there’s an artificial time pressure involved. That said, my experience is that LLMs usually get front-end code right more often - or at least not nearly as wrong - and it’s usually easier to fix up, too.

A cartoon bunny is rubbing a box cheese grater against its head, and bits are falling out of the bottom. The caption reads: Trying desperately to get that image out of my head.

I am now angry with myself for continuing to scroll through my timeline. I think I’ll take the rest of the day off from social media.

This is the exact problem I foresee. It’s not safe, and others’ livelihoods (perhaps even their lives) depend on what I release. If there’s a mistake in my software, I want it to be an honest mistake, not one caused by laziness or artificial time pressure.

Yes. This. Please.

You don't need to use the industry's marketing terms when informing your readers, because those terms are actively misinforming them.

An LLM is a good front-end to an AI system. That's all it can be, though. A true expert system is a bunch of different models working around a good reasoning model. The real problem with an LLM is that it has necessarily primitive reasoning and learning models, and in that sense, yes, they are bad.

What most people would want was known in the olden days as an expert system. We have very few of these, and they tend to be highly specialized to their particular application for hopefully obvious reasons.

Nah. If you want what people are asking of these models, you need the rest of the stack - not just the LLM, but also an extensive knowledge corpus, a full reasoning model, and several other components. These companies aren’t doing any of that because they’re very hard to do.

LLMs aren’t reasoning models, though. They use a primitive form of reasoning to understand the request, but their primary purpose is advanced pattern-matching. They also don’t usually have an extensive knowledge corpus, either - mostly just unvetted prior inputs.

The need for better and more precisely used terminology is one of my AI hobbyhorses. The trouble with "reasoning" is that it's definitely the wrong word, there needs to be a better word to describe that class of models, but I can't think of what that bettter term should be.

I don’t trust anyone who writes software with an LLM. They get things subtly wrong too often and write terrible code. I’ll use it to give me a rough idea, maybe (I’d rather read the documentation) and figure out what inputs break my code, but that’s it.

Ritalin, a drug so addictive they had to invent an extended release version because people kept forgetting to take their second dose

The sheer number of times I’ve been told I need to “try harder” to focus and stay on-task and not let my serial obsessions distract me… this *is* me trying harder!

No kidding. I’ve taken many a nap while on 30 mg of Adderall XR - plus coffee and an energy drink. If anything, I only felt more relaxed!

THREAD: Judge Ellis is the first federal judge to review extensive body cam video of DHS's actions in Chicago. She finds that DHS *repeatedly* misled the public and made claims that were disproven by agents' own videos.

I'll go through some of the most egregious ones here.