Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
martin-t
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
59 ms
·
751.
▲
by
martin-t
2y ago
Honest question: why do you censor the word rape? I skimmed HN's FAQ and I see no mention of forbidden topics or words and it seems the mod team prides itself on openness (does not make it true, i know). Do you believe your comment or
752.
▲
by
martin-t
2y ago
I am increasingly convinced we need to separate content storage from recommendation algorithms. When an organization does both, it has perverse incentives such as maximizing add-profit or user-engagement by building echo chambers instead of
753.
▲
Fascists and Rakes (2014)
(reasonableapproximation.net)
46 points
by
martin-t
2y ago
|
60 comments
754.
▲
by
martin-t
2y ago
This is what most people don't understand or don't want to understand. If the value proposition for antisocial behavior is positive when successful and zero when caught, then antisocial people will keep trying. Self-policing or te
755.
▲
Sally–Anne Test
(en.wikipedia.org)
1 points
by
martin-t
2y ago
|
0 comments
756.
▲
by
martin-t
2y ago
I wish cooperatives were more common and that society as a whole had more organizational memory and experience how to scale them.
757.
▲
by
martin-t
2y ago
I love The Expanse and it gets things right more than other sci-fi. However, I think it vastly _underestimates_ the amount of injustice than can be caused by powerful people with the help of advanced technology and ML. 1) You can literally
758.
▲
by
martin-t
2y ago
> how you think it’ll be our doom There's 2 main possibilities: 1) Self aware AI with its own agency / free will / goals. This is much harder to predict and is IMO less likely with the current approaches so i'll skip
759.
▲
by
martin-t
2y ago
> You clearly think Bevy and ECS as a choice for game development needs to be heavily motivated I think if there are 2 approaches people should know both exist and if both are roughly 50:50 in terms of capabilities, then people should be
760.
▲
by
martin-t
2y ago
On reddit i clearly say that ECS has its uses, such as large teams or games which need to add and remove components often. Most games people write in rust don't fall into those categories. Most people in rust have never heard of genera
761.
▲
by
martin-t
2y ago
You can't mix and match, you're still tied to bevy_ecs if you use any level of abstraction above it. I also didn't say it was a bad things. I just pointed out that everyone's goto example of bevy's commercial succ
762.
▲
by
martin-t
2y ago
Isn't Tinyglade mostly custom code including the renderer, just using bevy_ecs as data storage?
763.
▲
by
martin-t
2y ago
He's doing it because he likes learning and coding. And he knows what he's doing. Rust has huge issues attracting experienced gamedevs (unlike any other area) for some reason and mrDIMAS is one of the few exception, originally fro
764.
▲
by
martin-t
2y ago
Yes, the other is Fyrox. I don't think there are other viable 3D engines in Rust but at this point the hype surrounding Bevy is so large people wouldn't notice (and I stopped following the scene for the most part). There are also
765.
▲
by
martin-t
2y ago
You have a point but I don't think it's black and white. There is a spectrum between full scale revolution / civil war and targeted assassinations. The number of guns is also a spectrum and Russia is on par with many European
766.
▲
by
martin-t
2y ago
I don't think it's a fallacy. As someone who has 2 WIP games in Rust, I noticed that getting a very minimal gameplay prototype was pretty easy (given a sane engine choice)* but then adding features made the code balloon in size. S
767.
▲
by
martin-t
2y ago
> How do you prove whether or not someone has used AI? In this case it's rather easy, they admit it officially and publicly. In other cases it might be harder, we might need a whistleblower. I am sure in some cases where it happened
768.
▲
by
martin-t
2y ago
> Indeed, you cannot prove which process was used after the fact, especially not with AI. Citation needed. > anyone could then allege some human work is in fact AI and hence cannot be copyrighted I didn't say that. I said the cop
769.
▲
by
martin-t
2y ago
They need equipment too, lots of it. The west for example sent only token amount of tanks. It's slightly better in some categories like towed artillery but not nearly enough. People who disagree simply don't understand the scale o
770.
▲
by
martin-t
2y ago
I feel you. I started programming because I enjoy understanding something deeply and building things with that understanding. I like to write code that works on the first try because that means my mental model is correct. If I end up writin
771.
▲
by
martin-t
2y ago
Let me reiterate, there is no AI, there are only statistical models. It doesn't matter if you make the math too complex for a human to make sense of or not, it's just a remix of existing copyrighted material. Your argument forgets
772.
▲
by
martin-t
2y ago
How does this even work? Is hugging face hosting just the weights or some custom code? If it's just weights then I don't see how it could error out, it's just math. Do these chinese models have extra code checking the output
773.
▲
by
martin-t
2y ago
They completely forgot a 4th option: it belongs to the people who have copyright on the training data. If I take someone's book, painstakingly replace all words with their synonyms and publish it as mine, I am getting sued. Same if I d
774.
▲
by
martin-t
2y ago
Just 3 months? That's deleting 800 lines a day, each day. Did you need to read through the original code, get a deep understanding and match its behavior exactly or did you throw away huge chunks and write new code as you thought it sh
775.
▲
by
martin-t
2y ago
I inherited 100k LoC "documented" by an LLM. LLMs can put into words what a programmer can infer from struct definitions, function signatures, names and even taking a cursory look at a function body. LLMs cannot retroactively find
776.
▲
by
martin-t
2y ago
Not gonna happen. AoC always starts with beginner level problems. That's why it's so commonly used for learning the basics of new languages. A problem that wouldn't be immediately solvable by LLMs would either be too advanced
777.
▲
by
martin-t
2y ago
A classic example of a community self regulating until overwhelmed at which point rules are imposed which bad previously accepted and harmless behavior. Rules must take scale into account and do it explicitly to avoid selective enforcement.
778.
▲
by
martin-t
2y ago
"If they're the owners, do they get all the profit?" When you know you're not gonna work somewhere, might as well have fun raising eyebrows. Cooperatives really should be more common.
779.
▲
by
martin-t
2y ago
I agree norms are a distraction. It's about incentives. Every group/organization has an incentive to look strong to outsiders. It therefore minimizes internal conflict because that makes it weak. It most importantly minimizes visi
780.
▲
by
martin-t
2y ago
The concern isn't Earth organisms evolving and attacking us. The concern is they'll make it harder to detect potential life on other planets (because of false positives) or even that they'll destroy alien life if it exists.
More ›