meltyness

hm, there's an effort but scoped only to the std

It's terrifically complex: https://doc.rust-lang.org/beta/nightly-rustc/rustc_middle/ty...

- dns cache configuration

- wlan power saving which is kind of aggressive since Linux has had some power management issues

It's really why I go for a minimal distro, debian with xfce. There's not many components, they're not "tuned", and I can just kinda research any issues myself and find what works. Usually there's some pretty big gaps (missing a whole component, wildly malfunctioning, high resource utilization) and then it's easy to figure out. There's not much code, or really hard scripts to understand. Just maybe enable debug on the service. Learn where the developers put diagnostics (about:support is a godsend). You get to architect your hardware's success and get a rock-solid system. But yeah, if it's weird it might be weird for a while, haha.

For my setup I think the last few lessons were like:

- use gamescope for X11 Wine/Proton (huge.)

- Firefox profile reset to fix hardware decoding (followed some good guides and some bad guides)

- to make changes use the same git tag as your distro so the deps are easily in-reach.

My favorite new discovery for debian in particular is their extensive functional docs. https://www.debian.org/doc/manuals/debian-reference/

absolutely adore this.

It's not interesting due to the fact that it suggests humans are still in the loop of some slow-cycle improvements. That'd never get by any board. In fact, selection of model modes implies it's your responsibility, so that meal was scraped into your flowerpot years ago.

I'd say fat chance.

  vocabulary*

  *In the code above, we collect all unique characters across the dataset

At that point it was a game of "I'm not slandering you" to chip away at every other valuation, that could have easily have just been called antitrust because they didn't build it. That was 1996-2005 and went completely unchecked.

This is similar but the stack was even cheaper, and closer to more people's faces.

Even if governments take no recourse, I don't see an issue with government using it's position to put a food pyramid in citizen's faces to say like, "this can be harmful." The church probably would have if this were long ago, except, instead of fire and brimstone, some sort of epic story of social isolation, permanent dissatisfaction, and self-imposed constraints, alien abduction, transformation into a pig by a wizard?

There's probably a lot of visceral fears that would be worthy analogs to the harms of the feed.

I don't think that this narrative has been explored enough, honestly. Corps keep building crap like this, even amazon has (had?) an influencer feed.

People who are in play/leisure should probably practice tolerating more choices than "express mild, momentary dissatisfaction and receive an instantaneous reward"... that's probably not a life everyone should be trained to live

https://github.com/moby/moby/commit/1cbdaebaa1c2326e57945333...

I've built ollama before too, but, I like that I can cleanly rip it out of my system or upgrade it without handing root off to some shell script somewhere I guess.

If anyone's gonna bash up my system it oughta be me

I was rigging this up, myself, and conciscious of the fact that basic docker is "all or none" for container port forwarding because it's for presenting network services, had to dig around with iptables so it'd be similar to binding on localhost.

The use case https://github.com/meltyness/tax-pal

The ollama container is fairly easy to deploy, and supports GPU inference through container toolkit. I'd imagine many of these are docker containers.

Is it likely that it's a bigger problem to try and apply qualitative policies to training data, activations, and outputs than the approach ML-guys think is primarily appropriate (ie., nn training) or is it a bigger problem to scale hardware and explore activation architectures that have more effective representation[0], and make a better model? If you go after the data but cascade a model in to rewrite history that's obviously going to be expensive, but easy. Going after outputs is cheap and easy but not terrifically effective... but do we leave the gears rusty? Probably we shouldn't.

It's obfuscation to assert that there's some greater policy that must be applied to models beyond the automatic modeling that happens, unless there's some specific outcome you intend to prevent, namely censorship at this point, maybe optimistically you can prevent it from lying? Such application of policies have primarily targeted solutions that reduce model efficacy and universality.

[0] https://news.ycombinator.com/item?id=35703367

I think current censorship capabilities can be surmounted with just the classic techniques; write a song that... x is y and y is z... express in base64, though stuff like, what gemmascope maybe can still find whole segments of activation?

It seems like a lot of energy to only make a system worse.

edit: I'm speculating here that the supply chain wasn't already state-side for these players without knowing much about their business model

It's uninteresting because it's basically become a platform for regulatory capture. It's a wellspring of obviously non-universal ideas like, "there is no right way to integrate AI and primary education", "the federal government should subsidize ai access", or "only safe ai platforms should be permitted". I mean it's obviously their right to blather incessantly about it, I just think it's boring, and that's all I've said.

Maybe it's because I'm not a politician or a philanthropist, and I'm not required to tailor my actions to appease a large number of people subject to my will, but there's obviously better ways to approach that, like delegating and talking to people, who are local to the concern.

It's a nuanced and long term discussion and I think lots of the stuff that winds up in these interviews is really a local issue that's going into the wrong channel by well-meaning folks who don't understand government, or worse folks who are seeking to exploit government for profit.

And concretely, the interview doesn't focus on the book or the study, it's literally just an authoritative "intersectional" quiz about how AI/Education crosses with Diversity, Equity, and Inclusion,... a dumb question.