Unable to load image

Yandex open sources 100B GPT-like model :marseysaluteussr::marseyrussiaglow:

https://old.reddit.com/r/programming/comments/vit8xs/yandex_open_sources_100b_gptlike_model?sort=controversial

PSA: Yandex is a multi-billion Moscow based company, finances the Russian war of aggression in Ukraine, and is one of the main Kremlin's tool in spreading propaganda and suppressing dissent.

:#soyjaktalking:

YaLM 100B is a GPT-like neural network for generating and processing text. It can be used freely by developers and researchers from all over the world.

The model leverages 100 billion parameters. It took 65 days to train the model on a cluster of 800 A100 graphics cards and 1.7 TB of online texts, books, and countless other sources in both English and Russian.

Training details and best practices on acceleration and stabilizations can be found on Medium (English) and Habr (Russian) articles.

Make sure to have 200GB of free disk space before downloading weights. The model (code is based on microsoft/DeepSpeedExamples/Megatron-LM-v1.1.5-ZeRO3) is supposed to run on multiple GPUs with tensor parallelism. It was tested on 4 (A100 80g) and 8 (V100 32g) GPUs, but is able to work with different configurations with ≈200GB of GPU memory in total which divide weight dimensions correctly (e.g. 16, 64, 128).

https://github.com/yandex/YaLM-100B

https://news.ycombinator.com/item?id=31846593

66
Jump in the discussion.

No email address required.

This is based, AI companies are gating access to AIs to prevent people from misusing them. Glad Russia is giving out a reasonably powered AI and telling people to have fun

Jump in the discussion.

No email address required.

"Misusing" them how exactly? Making them say BIPOC?

Jump in the discussion.

No email address required.

A bunch of wet blanket "AI ethicists" (ie. foids who can't code but want to nanny an emerging field) are worried that being able to fake human interaction at scale/deepfakes will harm society. They are r-slurred though.

Jump in the discussion.

No email address required.

Tbf, the idea of social media platforms shadowbanning people by putting them in a closed ecosystem populated entirely by AI so they can't tell is concerning. A future full of people even more atomized and isolated because our elites have replaced even online interactions with a simulacrum designed to keep people docile and controllable.

Or even more effective astroturfing being used to shill and create a fake consensus. It's already happening, and imagining how much better it could get is concerning.

It's not really a genie we can keep in the bottle though. Keeping it out of the general public's hands is less of a concern than what governments and corpos will do with it.

Jump in the discussion.

No email address required.

And the longer you lock the tech away, the stronger will the first attack be against a naive population when AI finally becomes open

Jump in the discussion.

No email address required.

agreed, but for different reasons. Languages models are 100% being used to dilute wrongthink forums with human-like gibberish, but ethicists are still r-slurred. They think that gatekeeping ai will stop it from being misused, when really all it does is stop the proles from participating in the fun (ironic since most of these types are leftists)

Heres proof that my :marseyleftoidschizo: idea is used IRL (even if only for shits and giggles)

Jump in the discussion.

No email address required.

they aren't wrong the problem is that the harm has already been done when the internet became accessible to the public.

Jump in the discussion.

No email address required.

"Only approved context setters are allowed to do that!!"

:#soyjaktalking:

Jump in the discussion.

No email address required.

AI companies are gating access to AIs to prevent people from misusing them.

They are doing it to stop anyone from having any fun.

Jump in the discussion.

No email address required.

Agreed, we should be able to do all sorts of fun trolling with AIs.

Jump in the discussion.

No email address required.

If trolling is like fishing, AI trolling is using hand grenades. So of course people who don't understand fishing will hate it.

Jump in the discussion.

No email address required.

:#karlmarxey:

Jump in the discussion.

No email address required.

even libgen is Russian right?

Jump in the discussion.

No email address required.

Link copied to clipboard
Action successful!
Error, please refresh the page and try again.