Comment on Scientists discover that feeding AI models 10% 4chan trash actually makes them better behaved
Mr_Dr_Oink@lemmy.world 4 days ago
So is it saying essentially that in order to not output garbage, it needs to know first what garbage is?
Is it just me that things this seems like a no-brainer?
It almosr draws parallels to many societal issues. Knowledge is power.
People tend towards intolerance and hatred when they dont understand the thing they are angry at. The more they know the better they behave.
Is it just me that things this seems like a no-brainer?
Yes, and no. When raising our children, my wife prefers the “ban the bad stuff” approach. I don’t encourage exposure to bad stuff, but when my kid wants to buy and watch a raunchy movie, instead of yelling “NO!” and making him put it back, I let him buy it and we watch it, together, pausing to explain the unrealistic and awful parts and explain how imitating these things in real life can cause problems for you.
halowpeano@lemmy.world 4 days ago
No it’s more of a technical discussion. Many people might believe that in order to avoid toxicity, you just train a model on “good” non-toxic data and then apply toxicity removal techniques to address emergent toxicity that the model might spit out. This paper is saying they found it more effective to train the model on a small percentage of “bad” toxic data on purpose, then apply those same toxicity removal techniques. For some reason, that actually generated less total toxicity. It’s an interesting result. A wild guess on my part, but I’m thinking training the model with toxic content “sharpened” the toxicity when it was generated, making it easier for those removal tools to identify it.
MangoCats@feddit.it 4 days ago
Toxicity is everywhere, you can’t recognize that “Drill baby drill” has sexual connotations if you’ve never been exposed to sexual double entendre like that before.