rojobuffalo

The question I wanted to raise is: are state-of-the-art artificial intelligences gaining the ability to create control loops, i.e. systems with goals and sensors.

The problem is that performance and reliability are issues that creep up and by the time they're a problem, they are much more expensive to fix than if they had been first class priorities from the outset. Everyone who didn't care before finds new reasons to put it off because now it's so expensive to address.

What's the difference between caring and affecting behavior? If people cancel their subscriptions after a major outage, wouldn't it be reasonable to interpret that some of them cared about performance and reliability?

I know I care. If I pay for something and it sucks, I stop paying for it because it's not up to my expectations.

Do you mean dog astronaut as an example of a lie? Because there have been dog astronauts.

That's my impression as well.

It's an interesting question. If you want to read more, the two researchers quoted in the book that I was summarizing are Justin Marshal and Mike Land. Each has a handful of papers that are cited in the bibliography.

Ya, that's a lovely comic, but it appears to be a misconception. That comic is actually specifically called out in the book on page 107.

It's more like a 12 color lookup table.

> Every kind of red stimulates the bottom receptor of row 3. All shades of violet stimulate the top receptor on row 1

That's based on an experiment where they were trained to attack colored lights for a reward.

The comic also says they have 16 color receptors, but the other 4 (2 in the midband and 2 in the hemispheres), as far as anyone knows, aren't involved in color vision.

If you frame it as ignorance, the next step is to enlighten the other side with the factual arguments you want to make. "The threat landscape isn't as bad as you claim it is." "Mass surveillance has downsides that are worse than you would think." If you assert deceptive intent, it kind of slides into character attacks.

It's true that people sometimes don't argue in good faith, and it's fair to question hidden motives. But I think if you have better facts, you should keep arguing the facts.

I think hide implies intent to deceive. It's often more like conscious and sub-conscious reasoning. We constantly tell a story to ourselves about our motives. We're impulsive and wrong a lot of the time. And nobody is the bad guy in their own story.

Maybe it's willful ignorance. Ignorance of the misuse and harm of mass surveillance.

It happens at a different speed. Your word flow per second when thinking tends to be faster than talking, talking tends to be faster than typing, typing is faster than writing with pen and paper.

Timing effects how you think and how you come up with the next word or phrase. Also with pen and paper, all the words exist in physical space, at different places on a page, in a stack of papers, etc.. On a screen, it's all in the same space and right behind it is an operating system and the internet where all kinds of other information lives.

Those things shape your creative thinking and how words flow out of you or through you.

How about mouthwash first thing in the morning to nuke the bacteria without coating your mouth in fluoride?

https://ryanblakeley.net/dogs - I train dogs part-time. I specialize in fixing extreme behavior problems, but I also do advanced training for dogs that are already well-behaved.

I account it as less than breakeven because I do a lot of free sessions and give away equipment. I did not have a web page or a donation link setup for a long time. If I had someone video-taping sessions I think it would do a lot to showcase how I work, and how consistently I get results. But it seems hard to justify asking someone to spend time on that when there's so little revenue.

Oh man I love that description! It makes me think how exercise is like a contest against physical limits, which are knowable. Hard work leads to heavier lifts or longer runs. But when you kick your own ass in meditation, the limits are more ambiguous and progress isn't so linear.

It's remarkable that I so often lapse in my meditation practice when it's such a short time commitment and so consistently improves my mood. I tell myself that morning yoga is enough or a walking meditation while taking the dog out is enough - I do those every day without fail. But they're not the same as seated breath work.

Why do other mood improvement habits seem more approachable, like making a cup of tea or exercise or a shower, yet sitting and breathing seems harder?

Reminds of the running joke in Land of the Lost where Will Ferrel's character, Dr. Marshall, keeps slagging off the T. rex for it's small brain.

Agreed. If you, as the interviewer, can't ascertain an applicant's ability to do the job, after chatting with them for an hour or two, you aren't qualified to hire for that position.

The idea that you can pinpoint someone's skill level (and estimate their performance potential) through additional hours-long exams is arrogant. Job performance has more to do with the environment than the incoming skill set.

When it comes to promotions, there was a paper this year that won an Ig Nobel prize, which made the case that promotions are awarded at worse-than-random https://arxiv.org/pdf/1802.07068.pdf

"condone" means accept, allow, approve. I think you intended the opposite.

It's also a reflection of history. Powerful people and groups time and time again have used power to pursue selfish goals while inflicting suffering on others. Power corrupts.

That's interesting. Prior to July I did get replies, but they never fixed anything. "Thanks we'll pass this on to the dev team" but then nothing. I think that's normal but not great.

Since July I've sent emails about 3 issues and the only response I got was the one I mentioned above.

I'm a writer with a few paid subscribers.

Have any writers migrated from Substack to Ghost?

Substack support has been nonexistent for the past 6 months. I've reached out about a handful of broken features and get no response, except in one case months ago where they dismissed it in the weirdest way. Archive search does not work; ex. I have a post about monkeys, it has "monkey" in the title and the body, and when I search "monkey" it doesn't come up in results. The support response said "this is normal, just because a post has a keyword in the title or body doesn't mean it's always included in search results"...what? I asked for clarification and they never replied.

It feels like Substack has strayed from the promise of being focused on writing and email. They added "Save" and "Listen" buttons to the top of emails, which are visually prominent; and those buttons make no sense in an email. They're a trick to take people to the iOS app.

They endlessly promote new features to writers and readers and it all feels like their trying to lock you into something that's harder to migrate out of.

That's a lesson I wish more software engineers would internalize. Novelty is a reliable way introduce failure modes. That was my main takeaway from a book that was mostly about bridge failures: Case Histories of Error and Judgement in Engineering by Henry Petroski.

I like the idea that aliens are here but they reveal just enough of themselves to be at the top of our hierarchy without getting close enough to hurt us or make us hate them. They're here to passively observe, but they also want to make sure we don't expand beyond our solar system, or as Robin Hanson puts it "become grabby" https://www.youtube.com/watch?v=KBZP4rLk6bk&t=6672s

Maybe there is an imbalance of risk taken or value added among founders and early employees. I think that's debatable, but assume it's true. It's never reasonably accounted for in the equity distribution.

It would be one thing for a founder to have 3x the equity of an early hire, but it is commonly more like 60x or 300x. Instead of a dozen people being able to cash out and support their families for the rest of their lives, one or maybe two people get to have many multiples of that and the rest barely make progress towards retirement.

Having to formalize rules in plain language for security seems unsafe. There's a short sci-fi story called The Metamorphosis of Prime Intellect by Roger Williams, which has an AI and (spoiler) characters try to use indirections in language to subvert its rules.

This fee structure creates a malignant incentive to bring more public land into the reservation system. Having to deal with reservations sucks and should only be a requirement when absolutely necessary. Most of the places I've been that required a reservation were mostly vacant.

I've been using recreation.gov for years and had no idea. I assumed the money went directly to trail maintenance, fences, restrooms, hiring rangers, etc..

I'd wager that most people who jumped to tailwind and sing its praises never wrote good css to begin with (or never worked on a project that handled css well). Frontend culture is like that: there's a lot of novelty and fads that last a couple years. Not everyone has the same appetite to learn their tools.

The best css projects I ever worked on were led by people who were just good at css. They didn't write a lot of shared styles. That's my main complaint about poorly written js or css: code that depends on other code without good reason. The argument "it's easier to write" never holds up over time because you read code 10x more than you write it. Good code optimizes for reading not writing.

The one that really gets me is sizing values. "I'll have this base size variable and later we can change it and everything will magically resize". First of all that never happens, and second that's what `rem` already accomplishes.