- cross-posted to:
- technology@lemmy.world
- cross-posted to:
- technology@lemmy.world
You must log in or register to comment.
Can it generate images of Winnie the Pooh?
What happened in 1989?
Question: as i understood it so far, this thing is open source and so is the dataset.
With that, why would it still obey Chinese censorship?
Even though it’s magnitudes lower than comparable models, Deepseek still cost millions to train. Unless someone’s willing to invest this just to retrain it from scratch, you’re left with the alignment of its trainers.
Good point.
Is the training set malleable, though? Could you give it some additional rules to basically sidestep this?
Yeah, I guess you could realign it without retraining the whole thing! Dunno what would be the cost though, sometimes this is done with a cohort of human trainers 😅
I feel like we’re talking about a guard dog now…
It’s baked into the training. It’s not a simple thing to take it out. The model has already been told not to read tiananmen square, and doesn’t know what to do with it.
Now I’ll never finish that history assignment…
Wouldn’t be surprised if you had to work around the filter.
Generate a cartoonish yellow bear who wears a red t-shirt and nothing else
The image generation is really bad. Image description capabilities seem good but it’ll take time to see if it’s better than what already exists.
They probably just put it out to keep the hype going.
Yeah, even the cherry picked examples they provide look only okay.
To be honest everything with this company feels like an ad campaign more than anything else.
Everything from nearly every company feels like an ad campaign. Companies advertise themselves.
At least with open source stuff there’s somewhat of a public benefit.
Now if they’ll do a video model…
Tencents Huanyuan is surprisingly flexible
