3 ms·

I wonder what they did to Grok. The bimodal distribution makes me think it was just a system prompt and not a completely different set of right-wing training da

by vjans 2mo ago

I wonder what they did to Grok. The bimodal distribution makes me think it was just a system prompt and not a completely different set of right-wing training data. A 4chan trained LLM would probably actually be auth right.

cwmoore 2mo ago
So is auth right alt-right with authority?
- UncleMeat 2mo ago
  The political compass (which is not scientific in any manner and is basically trash for understanding things) splits into two dimensions. Liberal-Authoritarian and Left-Right. "Auth-right" refers to somebody in the Authoritarian/Right quadrant. Online, self-professed "auth-right" people tend to be fascists or fascist-adjacent and would get along well with a typical poster on /pol/.
sillysaurusx 2mo ago
Perhaps, but the last time someone tried training on 4chan data, Twitter sent a hate mob after them. https://cset.georgetown.edu/newsletter/june-16-2022/ https://cset.georgetown.edu/newsletter/june-16-2022/ https://www.theverge.com/2022/6/8/23159465/youtuber-ai-bot-pol-gpt-4chan-yannic-kilcher-ethics https://www.theverge.com/2022/6/8/23159465/youtuber-ai-bot-p... You'd need to source your compute from someone who doesn't care about such things.
spott 2mo ago
Probably just not post-trained well enough. I think the argument that pre-training gives a lib-left model is probably correct, so Grok, pre-post-training has a lib-left bend, then they try and post-train it. I imagine it is hard to post-train a political bend, as it is wide ranging and touches so much. If they didn't give the resources to do it well, we could get something like this. I could also see people half-assing the training. If the system prompt is being ignored half the time then the model is worse than I thought, cause that is pretty bad.
solumos 2mo ago
Is there a reason you think 4chan would be auth right rather than lib right?

3 ms·

I wonder what they did to Grok. The bimodal distribution makes me think it was just a system prompt and not a completely different set of right-wing training da

by vjans 2mo ago

cwmoore 2mo ago
So is auth right alt-right with authority?
- UncleMeat 2mo ago
  The political compass (which is not scientific in any manner and is basically trash for understanding things) splits into two dimensions. Liberal-Authoritarian and Left-Right. "Auth-right" refers to somebody in the Authoritarian/Right quadrant. Online, self-professed "auth-right" people tend to be fascists or fascist-adjacent and would get along well with a typical poster on /pol/.
sillysaurusx 2mo ago
Perhaps, but the last time someone tried training on 4chan data, Twitter sent a hate mob after them. https://cset.georgetown.edu/newsletter/june-16-2022/ https://cset.georgetown.edu/newsletter/june-16-2022/ https://www.theverge.com/2022/6/8/23159465/youtuber-ai-bot-pol-gpt-4chan-yannic-kilcher-ethics https://www.theverge.com/2022/6/8/23159465/youtuber-ai-bot-p... You'd need to source your compute from someone who doesn't care about such things.
spott 2mo ago
Probably just not post-trained well enough. I think the argument that pre-training gives a lib-left model is probably correct, so Grok, pre-post-training has a lib-left bend, then they try and post-train it. I imagine it is hard to post-train a political bend, as it is wide ranging and touches so much. If they didn't give the resources to do it well, we could get something like this. I could also see people half-assing the training. If the system prompt is being ignored half the time then the model is worse than I thought, cause that is pretty bad.
solumos 2mo ago
Is there a reason you think 4chan would be auth right rather than lib right?