7 ms·
Considering how poor it is compared to other models, it really emphasises how important fine tuning is. Models with MUCH smaller parameter counts are outperfor
by swalsh 3y ago
Considering how poor it is compared to other models, it really emphasises how important fine tuning is. Models with MUCH smaller parameter counts are outperforming it in many metrics.
- gordian-mind 3y agoCurrent metrics are a poor way to measure the usefulness of LLMs.
- make3 3y agono it empathizes the importance of training smaller models for longer, like the Mistral "overtrained" models
- gdiamos 3y agoShow the proof? Does it include IFT?
- lukan 3y ago"it really emphasises how important fine tuning is" Or rather the quality of the training data?
- jakderrida 3y agoAren't they usually built on most of the same training data?
- fragmede 3y agothat's a subtle dig at the fact that they have all of Twitter as a training corpus to use, but we don't know how they weight tweets. which, we know they're not gonna be weighted evenly.
- rezonant 3y agoI'm sure just like in X's algorithms, @elon tweets are weighted heavily.
- nonethewiser 3y ago> I'm sure just like in X's algorithms, @elon tweets are weighted heavily. Are you sure or is it the literal opposite and you’re just speculating?
- convery 3y agoThe X algorithm is also opensource, so you can verify before commenting..
- threeseed 3y agoX algorithm Github project hasn't been updated in 8 months: https://github.com/twitter/the-algorithm https://github.com/twitter/the-algorithm So clearly they aren't running it in production. Also they didn't open source the list of people who are being artificially boosted e.g. Elon.
- fragmede 3y agojust because they open sourced it doesn't mean that's actually what they're running on it though
- chrisco255 3y agoIt's not like he needs boosting, he was one of Twitter's top followed accounts long before he bought them. He's pretty good at getting attention.
- latexr 3y agoAnd yet it’s not enough to curb the desire to tip the scales. https://arstechnica.com/tech-policy/2023/02/report-musk-had-twitter-engineers-boost-his-tweets-after-biden-got-more-views/ https://arstechnica.com/tech-policy/2023/02/report-musk-had-...
- lukan 3y agoNo idea about the current state, but the open sourcing did show, they were favoring elon: https://mashable.com/article/twitter-releases-algorithm-showing-it-tracks-elon-musk-tweets https://mashable.com/article/twitter-releases-algorithm-show... And personally I never used Twitter much, but I certainly did not follow Elon Musk when I did - yet I had to see lot's of his posts in my feed. Surely just coincidence.
- llm_trw 3y agoWe don't know since no one is releasing their data. Calling these models open source is like calling a binary open source because you can download it. Which in this day and age isn't far from where were at.
- drexlspivey 3y agoTheir data is the twitter corpus which is public. Or do you want a dump of their database for free too?
- jahsome 3y ago[flagged]
- drexlspivey 3y agoRequiring a company to publish their production database for free is delusional. I haven't mentioned musk anywhere in my comment, you must be obsessed with him.
- jahsome 3y agoIt's fascinating you doubled down on your own straw man and still have the nerve to call others delusional. You missed my point, which I wasn't very clear about, so my mistake. Although it doesn't seem like you're interested in understanding anyway.
- llm_trw 3y agoSaying "It's just the twitter public corpus." is like saying "Here's the Linux Kernel, makefiles not included."
- zx8080 3y agoOr even "here's the Linux Kernel makefiles, no sources included, enjoy".
- GaggiX 3y agoOr even how much it was trained on this dataset, the amount of FLOPs.
- lairv 3y agoI would say it emphasises that training a good model is more than throwing random data and compute