Patching a newer version of the Youtube app resolved the issues with playback I was having.

Not if Anna has anything to say about it…

The good ol’ Linus parrots. Squawk “Steve Burke is a bad journalist because he pointed out errors publicly that affected consumers.” Screech “Linus didn’t sell the employees internally on the idea that he and his wife were a substitute for HR, he auctioned it.”

More often than not, people who are passionate about something, such as Linux, take personal offense when someone says something incorrect or offensive about said thing. Oh, and blud is just to call someone a poser.

I love this comment so much. “You crossed Linux? Now you’ve crossed me, blud.”

To be fair, the comments and posts you leave are technically being collected for display across the lemmyverse. In that sense, there’s never going to be a zero data collection Lemmy client. Still, Liftoff currently has my vote. A decent little FOSS fork of Lemur, I believe.

Heck, even my college Sociology textbook from OpenStax basically has nuclear fear-mongering baked into one of the later sections.

Unfortunately, there’s still that one guy in the comments trying to say that hypothetical, largely unproven solutions are better for baseload than something that’s worked for decades.

Yeah, that’s fair. The early versions GPT3 kinda sucked compared to what we have now. For example, it basically couldn’t rhyme. RLHF or some of the more recent advanced seemed to turbocharge that aspect of LLMs.

So a few tidbits you reminded me of:

• You’re absolutely right: there’s what’s called an alignment problem between what the human thinks looks superficially like a quality answer and what would actually be a quality answer.

• You’re correct in that it will always be somewhat of an arms race to detect generated content, as lossy compression and metadata scrubbing can do a lot to make an image unrecognizable to detectors. A few people are trying to create some sort of integrity check for media files, but it would create more privacy issues than it would solve.

• We’ve had LLMs for quite some time now. I think the most notable release in recent history, aside from ChatGPT, was GPT2 in 2019, as it introduced a lot of people to to the concept. It was one of the first language models that was truly “large,” although they’ve gotten much bigger since the release of GPT3 in 2020. RLHF and the focus on fine-tuning for chat and instructability wasn’t really a thing until the past year.

• Retraining image models on generated imagery does seem to cause problems, but I’ve noticed fewer issues when people have trained FOSS LLMs on text from OpenAI. In fact, it seems to be a relatively popular way to build training or fine-tuning datasets. Perhaps training a model from scratch could present issues, but generally speaking, training a new model on generated text seems to be less of a problem.

• Critical reading and thinking was always a requirement, as I believe you say, but certainly it’s something needed for interpreting the output of LLMs in a factual context. I don’t really see LLMs themselves outperforming humans on reasoning at this stage, but the text they generate certainly will make those human traits more of a necessity.

• Most of the text models released by OpenAI are so-called “Generative Pretrained Transformer” models, with the keyword being “transformer.” Transformers are a separate model architecture from GANs, but are certainly similar in more than a few ways.

Unless I’m mistaken, aren’t GANs mostly old news? Most of the current SOTA image generation models and LLMs are either diffusion-based, transformers, or both. GANs can still generate some pretty darn impressive images, even from a few years ago, but they proved hard to steer and were often trained to generate a single kind of image.

I was incorrect; the first part of my answer was my initial guess, in which I thought a boolean was returned; this is not explicitly the case. I checked and found what you were saying in the second part of my answer.

You could use strict equality operators in a conditional to verify types before the main condition, or use Typescript if that’s your thing. Types are cool and great and important for a lot of scenarios (used them both in Java and Python), but I rarely run into issues with the script-level stuff I make in JavaScript.

If I remember correctly, 0 and 1 are considered falsy and truthy respectively, so it should be falsy and truthy and false which I believe would return false.

Tried it out to double-check, and the type of the first in the sequence is what ultimately is returned. It would still function the same way if you used it in a conditional, due to truthy/falsy values.

JS is one of the most fun programming languages ever created; how dare you slander its great name.

Tbh, I haven’t really had this issue in a few weeks. I’m tempted to think it’s usage-related, and could possibly indicate that my memory allocation for the DB is still too high.

Maybe Australia’s offerings are different, but I see this board with an 11th gen i5 for USD $299. There’s a ton in the $300-500 price range with several different configurations. That’s really the interesting range for doing hobby projects.

Well, framework has one cool side-effect of their repair-friendly approach: their laptop mainboard can be used as an SBC. I’ve seen a few projects use it in this way, and I believe they even sell an official plastic case for it. It’s a well-documented piece of computer hardware that is regularly refreshed and can be fitted easily into slim chassis.

Oh, and another cool thing is that their screens have magnetic bezels. ThinkPads are a PITA to fix if you just want to replace an LCD panel; framework makes it trivial to keep the upper chassis and only replace the part that’s actually broken. That’s the real pitch with Framework: replace anything easily and upgrade your computer for only the cost of the mainboard or socketable component. Some of their newer devices have a socketable PCIe expansion bay, which could be used for things like socketable GPU upgrades.

What would be pretty awesome is if Lemmy implemented an optional awards feature that could accept donations into a tip jar of some sort. Many of us running smaller instances appreciate selfless donations, but this could be a decent way to make supporting hosting costs more fun.

Like I said, I’m aware of extant measures to try and steer models, but people often assume a level of craftsmanship in censoring models that simply does not exist. Jailbreakchat.com is an endless stream of examples of this very fect; it’s very hard, especially with the limited context lengths of current models, to effectively give them any hard directives.

And back to foundational models, which are essentially free of censorship, they will still exhibit a similar level of political bias unless prompted otherwise. All this to say that, discounting OpenAI’s attempts to control their models, the model itself will inherently learn from and mirror the real-world biases of the text it was trained on. Those biases happen to fall along lines that often ignore subtlety in debates regarding illegality and morality.

It’s hard to say what LLMs are “programmed” to do, as they’re largely untamed beasts of text prediction. In fact, I would suspect its built-in biases are less the result of pre-prompting or post-foundational-model training and really just what a lot of people tend to think online. In a way, it’s more like people in general often equate illegality with immorality.

You can see similar biases in many of the open-source LLMs that are floating around. Even though they’re basically built outside of large corporate cultures and large-scale monetary incentive, they still retain a lot of political bias that tends to favor governmental measures heavily.