Another researcher says OpenAI trained on conversations, then claimed breakthrou

Posted by ColinWright 2 days ago

Counter262Comment15OpenOriginal

Comments

Comment by dang 2 days ago

Comments moved to https://news.ycombinator.com/item?id=49639408, which has the original source. I'll try to assemble the other links there too.

Comment by jarofgreen 2 days ago

Also discussed here: https://news.ycombinator.com/item?id=49638353

Original posts

https://mathstodon.xyz/@andreasthom/117240535270608201

https://mathstodon.xyz/@andreasthom/117240536885387540

https://mathstodon.xyz/@andreasthom/117240537520615623

This story is someone on bluesky screenshotting someone on twitter without a link. The person on twitter screenshotted the source on the fediverse without a link. This is getting daft.

Comment by netghost 2 days ago

And here we are on HN discussing it.

Comment by jarofgreen 2 days ago

Discussion is fine, let's just try actually linking back to the source.

Comment by lubujackson 2 days ago

Screenshotted this thread; will repost on Reddit and 9gag for the updoodles.

Comment by trinsic2 2 days ago

LOL Lets screenshot each others comments and comment on them.

Comment by pluc 2 days ago

You can pretty accurately fake a video these days, faking a screenshot is 7 year old level of skills.

Comment by michaelmrose 2 days ago

[dead]

Comment by zerozerotwo 2 days ago

A toot embedded in a tweet .

Comment by egypturnash 2 days ago

A toot in a tweet in a... skeet? Is there any consensus on what to call "a post on bluesky"?

Comment by Qworg 2 days ago

Skeets is what the community wants, but the company does not (for obvious reasons).

Comment by keeda 2 days ago

Bleat?

Comment by WhrRTheBaboons 2 days ago

flutter? idk

Comment by ColinWright 2 days ago

Comment by ChrisArchitect 2 days ago

Comment by vatsachak 2 days ago

Nah the models are just that good at short scale math now. If your research program is close to it's goal the model can get it.