Hacker News Reader

QALC is a vscode extension that behaves quite similarly to this!

infinityioon Show HN: Obsidian Canvas – An infinite space for your ideas

The difference is with a local-first editor (like Obsidian) you hopefully get to keep your files if the company stops being nice

infinityioon Promotion of alternative social platforms policy

To be fair - it looks like twitter is also banning links in bio here?

infinityioon Laion-5B: A new era of open large-scale multi-modal datasets

Obviously there would be limits as to how much could be manually reviewed by hand (if 1000 people reviewed 1000 images each, only 0.02% of the images would be reviewed assuming no overlap was required), but I wonder if there would be any benefit to attempting to crowdsource captions for the dataset for the worst available images

infinityioon Ask HN: How does ChatGPT work?

As far as I know Davinci is supposedly a variant of gpt3-175B, so likely 175B parameters, so likely in the hundreds of gigabytes of ram unfortunately

infinityioon AI will dominate the animation industry in less than 5 years

Three years ago, our SOTA image generation models had very strong limitations (when compared to today's models), it can be difficult to see when and where the next breakthroughs will come - NeRFs are not there yet, but we may not be far away

infinityioon Where has the passive radar code gone?

https://gist.github.com/rietta/60b7b3f7ca33bd13948c

infinityioon YouTube confirms that it has removed the “sort by oldest/newest” option

To my knowledge Amazon (via Twitch), Meta (via Facebook), and Microsoft (live streaming only, via Mixer) have all played with user-generated video with limited success

infinityioon 3D map of the Czech Republic, made from 12.5cm/px imagery (2017)

> Can this map help me find the town hall of Whittier, Alaska?

For those who (like me) who didn't get the reference and were curious about this - the town appears to be mainly consolidated into a single 14-story building [0]

[0] https://en.wikipedia.org/wiki/Begich_Towers

infinityioon Tesla engineers were on-site to evaluate the Twitter staff’s code, workers said

he may have been rich from his family, but he became Rich from zip2 and x.com/paypal (and subsequent re-investments)

infinityioon Largest score declines in NAEP mathematics at grades 4 and 8

From this specific data, this viewpoint is hard to justify, because the scores were at or near their highest in 2019, and the data shows modest improvements since the 90s

infinityioon Pocket Casts Mobile Apps Are Now Open Source

fyi - Pocket Casts is now part of Automatic, not neccesarily a start-up still

infinityioon Obsidian 1.0 – Personal knowledge base app

> Being able to sync for free (I just pay for S3 costs) means I got cross-device/platform support and know I will never lose my notes.

This is also the case with Obsidian! Because the files are just loose `.md` files in a folder, I personally use Syncthing and Git to keep them synchronised across devices for free. Sync is another option to complement those, for people who don't want to roll their own solution.

By default in Obsidian you own the data (and it never leaves your hard drive), you pay for them to host it for you so that it 'just works' across devices

infinityioon Google has added ads on both its search page and Chrome://newtab

To my knowledge, Google has 90% of search engine traffic in some regions, and Chrome makes up the majority of Web browsers, even disregarding Chromium

infinityioon The road to Zettalinux

Not quite - 1024GiB is 1TiB, so it's 1024 x 1024 x 16TiB drives

infinityioon S6: A standalone JIT compiler library for CPython

numba can now also support ahead-of-time compilation, in some cases

infinityioon F4PGA: Open FPGA Tooling: Xilinx 7-Series, Lattice iCE40/ECP5, QuickLogic EOS S3

I think the critical difference is that the 'emulation' is effectively done in hardware instead of software, so designs can be clocked much higher than traditional emulation (GHz+ speeds)

infinityioon Search 5.8B images used to train popular AI art models

Obviously it might be a slightly futile task given the size of the dataset, but would there be value in adding community captions to some of the images? I'd be happy to spend 15 minutes captioning the worst-labelled (maybe cosine similarity between stabdiff im2txt and the prompt?) images in the set / reviewing other people's captions, and if you can get enough people on board you could probably get through a not-insignificant number of new captions. Equally a risk of anti-ai groups sabotaging this process though

infinityioon YouTubers have to declare ads. Why doesn't anyone else? [video]

(posted Feb 2021)

infinityioon Shortest URLs on the Internet

unfortunately, it seems like ICANN rules prevent this for new gTLDs

infinityioon Bitwarden raises $100M

do you mean $30/user?

infinityioon DIY mechanical keyboard switch lets you set its actuation point

Recently, a few brands (Wooting, Razer, Steelseries, etc) have introduced keyboards with analogue-capable input (using either hall-effect, like the switch in the article, or optical sensors). This already allows the actuation to be set per-key in software, so your idea probably wouldn't even be that hard to implement with today's hardware! (although you may need a beefier microcontroller on the board)

infinityioon Bitwarden raises $100M

> But somehow I never got the TOTP working.

I think bitwarden's TOTP generation is a paid feature, but it still provides the option to store the secrets in the free version

infinityioon Instagram removes Pornhub's account

My interpretation was that they are rooting for mindgeek (being able to fix the numerous issues they have), because the alternative will be so much worse

infinityioon US Government Bans Export of Nvidia A100 and H100 GPUs to China and Russia

State-of-the-art chip fabrication is still mostly based in Taiwan, with TSMC (as an example: the BR100, China's most recent native flagship gpu, is fabbed using TSMC) - China's own SMIC has appeared to struggle

However, they seem to have been catching up fairly quickly - until very recently SMIC have been stuck around 14nm (state of the art in 2013, SMIC in 2019), but they recently announced they have managed to start producing 7nm chips (state of the art in 2018)

infinityioon Run Stable Diffusion on Intel CPUs

I've had success with GPT-J (6B) [0] and GPT-NeoX (20B) [1], but they probably aren't quite the quality level you'll want to have

On the other hand, Facebook has recently released the weights for a few sizes of their OPT model [2]. I haven't tried it, but that might be worth looking into, because they claim that their model is comparable to Davinci

Note that for CPU inference you will be unable to use float16 datatypes, otherwise it might error out

[0] https://huggingface.co/EleutherAI/gpt-j-6B [1] https://huggingface.co/EleutherAI/gpt-neox-20b [2] https://huggingface.co/facebook/opt-66b

infinityioon Relax for the same result (2015)

I think code blocks might allow for it, at least in clients that apply formatting?

infinityioon The Jupyter+Git problem is now solved

not a complete solution, but PyCharm and VSCode both support using `# %%` to split a python script into 'cells' (stolen from matlab?), which then be executed individually/repeatedly