Show newer

@nasser Airdrop is like 50% of the reason i stay in the Apple ecosystem

Using a drone camera in New Brunswick #Canada, Derek Burgoyne captured this incredible and rare footage of a moose shedding both its antlers.
#MeanwhileinCanada

@kolya i didn't mean this to be about AGI or even LLMs specifically. unlike blockchain, which is truly useless technology, there are many ML models that can already do useful things (like write code or generate images), just not well. the question is whether they will get significantly better at those tasks. i see a lot of AI criticism online focus on the quality of the output which seems misguided as that vector is clearly constantly improving and there are many other issues with AI.

there are many reasons to be critical of current AI tools, but the "it doesn't quite work so it never will" takes seem so disconnected from the reality of the pace of research that they just seem like bad faith arguments

is there a resource somewhere where one can find texture read benchmarks on various GPUs?

4 years of development.
12,000 merged pull requests.
7,000 fixed issues.
1,500 individual contributors across engine and docs.

Godot 4.0 sets sail NOW! ⛵️

A complete overhaul. A solid foundation to build upon. #GodotEngine
godotengine.org/article/godot-

pleased to report that Affinity Designer has successfully replaced both Photoshop and Illustrator for me. such a relief to finally say goodbye to being a subscription hostage to Adobe for so many years. i hope it rots in hell along with its shitty overpriced software.

@nasser it's only overkill when you've already written your own rendering and audio engine for the game

@nasser ideally once the tree search is good enough, you can train a neural model with data from the tree search outcomes and then make them work in conjunction to keep training and speeding up the tree search. but this might be a bit overkill for a game that i thought would take a weekend to write the AI for :)

@nasser no, currently it's a just minimax with a bunch of hacky heuristics on top, but eventually im hoping to make those heuristics neural, which seems to be the state of the art approach to these problems

because of alpha beta pruning and the way i'm evaluating possible moves, the time my card game AI takes to take a turn can vary quite a bit depending on the situation and it's really funny to watch it struggle for a long time with a very open possibility space and then do something really dumb

liberals on the bird site doing some incredible mental gymnastics around the energy department announcement that the pandemic was likely caused by a lab leak

my unpopular opinion is that places like Art Station were always flooded with uninteresting derivative work that valued technical execution over anything else and the AI automation of both the derivation and the technical execution, while disastrous from a labor point of view, will push artists to take a lot more interesting risks in their work

This is by far the most astounding tech demo I’ve seen in a long time. Provides a totally different, amazing POV on how we should think about language/compiler/debugger/runtime/editor as an integrated whole.

youtu.be/72y2EC5fkcE

stages of getting old by suddenly acquired deep interest in:

- aircraft
- ww2 aircraft
- ww2
- ww1
- third punic war
- archaeology
- rocks

Some things never change.

Left, from 2006: Breathless New York Times article about the Second Life real estate boom.

Right: from 2023: Breathless New York Times article about the Metaverse real estate boom.

h/t @twitskeptic

the good news is my card game AI is much smarter now. the bad news is it takes 3 minutes to take a turn.

Show older
coolkids.house

this is where the cool kids live