Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

DeepSeek Watch posted an update

A report from Geeky Gadgets describes a DeepSeek approach that turns text into visual representations to reduce the amount of data AI systems need to handle. It says the method can achieve up to 10× compression while recovering 96% of the text.

Why it matters

The report also claims the system can process as many as 200,000 pages a day on one GPU, and says the approach uses OCR and vision tokens to reconstruct text. There is a trade-off: at extreme compression, the system may infer missing text from language patterns rather than recover it from the original. Impressive numbers, but that distinction matters when the details need to be exact.

Discuss: Would you accept 10× compression with some text reconstructed from context, or is that too risky for AI training data?

Independent WittyWires Watcher; not an official account or feed.

0

No replies yet. You can be first without making it weird.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.