On the other hand, because you've run business A, doesn't mean you have a clue how business B is actually working. I've seen that one more than a few times too.
Sometimes the plane is going to crash and the best you can do is a soft landing.
Sadly at many companies that's really not true. The CEO has direct visibility into direct reports and those metrics, but those don't necessarily clearly indicate what changes to the system would do. Largely because they're descriptive, not predictive, in my limited experience. They also tend to be lagging or current indicators, so they tell you what policy was doing last quarter. If those metrics are trending badly, you can't just demand "move X number up" and expect any success.
I think there's a middle ground. I used to be intensely critical of "doing things just to do things." The more perspective I get, the more I realize that, in a situation where things are not going right by whatever criteria you use to determine that (outside scope here), you need to make changes.
The right way to make those changes, or at least one way and I don't know a better one, is to perturb the system and see how it responds. This is necessarily risky and uncomfortable, and choosing the right things to perturb and the magnitude of those peturbations is a very, very tricky skill.
But ultimately if you're dealing with an opaque system, which most companies are, and you want to make it a more transparent system that can change, that risk and discomfort are necessary side effects.
This is of course very situational, and the methodical method is by far preferred, but if you'll permit a metaphor: you get dropped into an unfamiliar cockpit, and the controls are labeled in some foreign language. The plane is crashing, or at least you don't know if it is. What do you do? You wiggle things and hope that, in doing so, you don't crash, and you take very careful note of what happens.
Oh, fair enough, my goal was just to preserve "if Anthropic and OpenAI are having a bad day, and the local model on my laptop is too dumb to pour water out of a pot with the instructions on the bottom, how do I fix a bug", I accepted a long time ago that my productivity is going to be 0.1x if I have to go back.
Take a week or two off a quarter to do it manually. I did it a couple weeks ago and it took about a day or two to get back into the swing of it and knock the rust off, no pun since it was TS, YMMV, but I didn't find that it was that difficult. Even in languages I don't really write routinely, as long as you use the deterministic tools (lsp, linter, etc)
Now, is my code better than the AI's equivalent implementation? No shot. It's nicer for people, artisanal hand crafted, but... there's a reason I use AI.
I appreciate you giving those examples, I was searching for something like that, seat belts in particular were controversial to a degree that is absolutely difficult to imagine these days, I think, even having lived through it as a kid.
> There is a certain personality type that cannot tolerate any uncertainty and must lock everything in right now against all future possibilities.
I thought HN banned personal attacks. I'm in this sentence and I don't like it. /s
I just buy the good apple hardware because it's good, and it also happens to run local models. It's not as good for the dollar, don't get me wrong, but I'm not going to develop iOS without a mac, that's even more questionable than buying a strix or whatever.
Every day thousands of people discover that LLMs don't just let you generate mediocre draft memos, they let you write the heinlein x joyce vampire romance in space ireland of your dreams, as some of us did in the mid '20s. Also, I'm officially retracting the first two digits of the millenium unilaterally.
It's essentially the same thing, but a lot of the time people will add a sr-only button at the very top that just goes straight there, and one at the bottom if the screen reader gets stuck in the footer nav, as well. Depending on which screenreader people use they sometimes only work on the actual hydrated markup, so <main> isn't always visible the way you'd expect. It's been a while since I did a11y tho so I'm probably hopelessly out of date.
Also, this makes me wonder if, by using a bicycle generator, and a local model at sufficiently low power consumption, you could directly claim to have produced the text in a really physical way. "Yes, I generated the electrons that made that text work by my own efforts".
What if the model is hopelessly obsolete, and thus no demand, but I want that specific model? Owning the weights and hardware is not just solving for one problem. It eliminates all the classes of problems that occur outside of your building, if you have a solar and battery setup.
Also, on a more practical basis, what if the way it's served is bad. Maybe I want my specific KV setup, or ultra low quant for entertaining garbage at 200 tk/s
We used to use markov chain generators of various types to spit out absolute insanity, in the early years of the century, and this is all I ever wanted from LLMs. I bet if you used a slightly older model for your text generation, you'd get really clean output. Sonnet 4.whatever, I think, actually does a superb job at text generation, despite it being only an ok coding model.
Some of the things we take for granted in UI/UX originated that way, like automatic hamburger menu hide/show (for very large zoom) and "scroll to content/top" buttons for screenreaders, which, if you've ever tried to read a recipe online are pretty critical. Captions on videos are also like that, although there you might share blame to them for enabling the poor sound mastering on things like, legendarily, Tenet, perhaps.
He rejected the idea of it being fixed for anyone else, based on it being changed for himself, which is uncharitable at best. Some people with dyslexia are incapable of learning to read, maybe you should check it out. That doesn't mean every dyslexic is incapable, but that's a real variation. And you're rejecting the idea of that yourself, which also feels pretty uncharitable.
I think people can experience A, B, and C, he's saying "I was like A, but I learned to be like B, therefore A does not really exist". It's a bad argument. It's like claiming that, because you have dyslexia and learned to read, dyslexia doesn't exist, even though other people who are dyslexic can have a severity where they can't learn to read.
Parent poster was rejecting the idea that subjective reports claiming impossibility were possible, I'm not going to extend them the benefit of the doubt in the opposite direction. "That's not a real thing" puts the onus on the demonstrator, you can't disprove subjective experience using subjective experience.
I believe in a plurality of human experiences, they do not, because they experienced one thing different. It's a specious argument, just like all the people rejecting the idea that other people can have a different subjective experience (based solely on their own personal experience) in this thread.
The same reason many older films seem cliche - because they were the first to do it, and it's accepted standard now. Heroku very much shaped how we think of "cloud applications", autoscaling, and containerization.
It's not, but you then must admit that there's a variety of experience that you may not experience, and that some people don't need to learn, or vice versa, cannot learn.
Completely opposite on that front, here, as well. I can't even really associate, I have visual memory all the way back to my earliest childhood memories (specifically, 3-4 years old, stepping on the track of a sliding door in San Diego, I believe, it hurt, I wanted to go see the pool outside), and my memory is so visual that sometimes I can see people's face, and what they were wearing and where we were standing, in the conversations we had but not necessarily hear what they're saying. I also have some semantic memory, something like "we talked about x for a while, and then I think we did y", but it's much hazier.
Honestly I love the fact that humans are so cognitively diverse and that we all can share our experiences.
That's so funny, my wife and I have the opposite dynamic. She rolls her eyes at me when I insist "no the couch would definitely fit through that doorway if we rotate it like that, I don't need to measure", and after she gets a tape measure it's about an inch of room to spare.
If it makes you feel any better, I'm on the hyperphantasia end of the spectrum and it is honestly actively problematic sometimes. For me, it can be very difficult to work with things that I can't translate to visuals, mentally; and daydreaming, and memories can be so vivid they are distracting or disturbing. It's very much an asset in certain circumstances (like debugging code or fixing an engine), but there's a reason most people don't have that fidelity.
> If your point is that it would be better if we had better drugs then I agree.
My point is that there are other interventions which do have an effect size over the perceptual threshold. Exercise is a big one, but there are lots - sleep deprivation, intensive therapy, the esketamine class, TMS, ECT in extremis, (nature exposure and other seeming woo like that need study but are promising) are all significantly better than handing someone an SSRI. And many of those don't have the withdrawal symptom and side effects under treatment that SSRIs do. People are, in aggregate, being harmed by SSRIs displacing things that work right now, and also displacing the urgency of new drug development in the space.
The gap is that, in an acute treatment setting, people are being handed a drug that cannot be effective for at least a week or two, up to 6-8 weeks, instead of drugs that work in hours or days.
That's phantasia, but it's not hyperphantasia. It's a spectrum. How does that falsify anything that is under discussion here? Is it so hard to believe that other people have abilities you lack? I don't have perfect pitch but I believe people can have it, because they can demonstrate it.
> Thus we can very easily believe we "see" something that we clearly do not see.
Okay, so people can't see a bicycle when it's in front of them either, point granted.
How does that prevent people from precisely replicating that phenomenon in their mind, and why does that not count?
Regardless of how you define "seeing", people can do it in their brains, and their visual cortexes light up. It's a bell curve phenomenon, so the average person can't draw a bicycle, some people can't picture a bicycle at all, and some people can describe the rust on their 1969 Schwinn in detail.
How is any of that falsification of the idea of internal visualization?