Finding spices under "supplements" sounds very British. /s

Best printer 2025: just buy a Brother laser printer, the winner is clear, middle finger in the air Best Printer 2025: wireless laser home office supply shop now Brother

I can't believe this ongoing series of articles by @reckless.bsky.social about which printer everyone should buy is my go-to recommendation for anyone for many years now.

www.theverge.com/tech/641940/...

At my first intern position back when I was in school, a customer with the last name "Panus" walked in. It was unique and got stuck in my head. Didn't want "anus" in the name so I changed it to "Panos". Later realized it means "the runs" in my mother tongue. Ended up being a butt joke nonetheless.

The blog post "Our agreement with the Department of War" post on the OpenAI website.

At the bottom of the blogpost, there's a small "Ask ChatGPT input". I typed "Is this blog post misleading or deceptive?"

ChatGPT admits the post is misleading.

It's nice that OpenAI includes a little "ask ChatGPT" thing at the bottom of their own bullshit so that even ChatGPT can make a statistical guess that it is indeed bullshit.

Gianmarco is talking about people who actually want to use them.

Steam wishlist button with the number 2009 on it, showing how many games are tracked by my wishlist

GGapp UI showing I tracked knowledge or progress on around 700 games via that platform.

And since then I've been recommending the game non-stop to friends and got two hooked on it recently. I think it's still my favorite small indie game of the recent... decade? (Especially for game devs!)
And that's saying something, since I know quite a lot of games. ^^'

Actually, I think your reasoning makes perfect sense. I think it's quite bad for LLMs if the training data can contain a thousand ways to achieve the same thing. It can probably lead to much more errors caused by >0 temperature settings, at least. And at =0, it would often generate copyrighted code!

I won't argue about Elon being that, I'm just thinking about the technical feasibility. If LLMs currently train on commented code, it's theoretically possible to translate the code into binary and attach said comments as annotations for the training data, no? (Then it could be language-agnostic.)

I mean, same as with other annotated training data, I assume. It will need to be annotated training data, though, of course.

I'm just not sure whether the omittance of variable names and of all the other overhead when programming would benefit the LLM or be detrimental.

That's a very valid point, actually, hm.
But also, it seems like it's technically "just" one more layer of training. I mean, technically an LLM could learn that a specific collection of ones and zeroes e.g. produces a matrix multiplication, no?

I don't think the richest man on earth, who builds his whole pride around that fact, would admit to anything that even resembles a flaw. (See his Jiu-jitsu beef with Zuckerberg a while back, lol)

That's not the same at all, though? LLMs can produce working code (most of the time). Yes, they can't do math, but that's why you tell them to write a calculator instead of telling them to add numbers. Coding in binary would probably need a similar type of training, I'd assume.

Home - 42 The Network 42 is a global, tuition-free tech school reinventing education through peer-to-peer learning and real-world, project-based experience.

@bennjordan.bsky.social idk if you read/see posts directed at you, but you were recently(?) talking about your education status and this reminded me of an interesting concept I came across recently. I'd really love to hear your opinion on it since I'm personally very torn: www.42network.org

We know he can't play games, though.
(But maybe doing that he'll learn.)

Just a thought: why is it nonsense? The "trick" with LLMs is that you spam them with so much content, they "learn" to predict what comes after. The needed context to predict stuff in binary is probably longer, but couldn't it theoretically work? Interacting with it would be impossible, though. ^^'

And you are rightfully proud of that! :D

If that is your view and you can actually work with Linux, this is fair. ¯\_(ツ)_/¯

An excerpt to the contrary of your second sentence. I agree with the paywall thing though (even though I didn't see it thanks to some plugin or browser setting).

Is there a point or argument somewhere in your comment?

This was not an argument. They just said they don't understand why it's mentioned as a plus, when it can be positive or negative, depending on the person.

You are absolutely incredible and I will never ever not stop doomscrolling just to marvel at every picture for minutes at a time.

WOAH that's actually insanely well done. :O

In other news: rain is wet.

But also: I wonder how many of those investors just KNOW many investors are dumb and planned accordingly. Had I stock in a AAA game company, I would've sold it after the announcement to buy it back at a lower price.

Wait, I thought Tibetan monks only exist in the DC franchise?
(jk)

I'm sorry for sounding so unnecessarily judgy. It really wasn't meant as an attack on you, I just developed an involuntary reaction to this movie, specifically. That, of course, is no reason to spread negativity under a post where someone shares something about themselves.

I feel like this fails the reverse of the Bechdel test. :'D

Can't offer much but this one thread I scrolled past a few minutes ago: bsky.app/profile/naur...
Although it mostly only provides more options instead of helping the choice. ^^'

I definitely agree with all other suggestions you had that I know of, but this one was a hard no for me. I couldn't even get through half of it because it was just so unnecessarily disturbing to me. ^^'
Although, to be fair, I knew of a twist in this movie before going in. Would've ruined it anyway.

This is a metroidvania boss if I've ever seen one