AI Training Data Bias
Discussions center on how biases in AI and machine learning models, particularly LLMs, originate from biased training data reflecting human and societal prejudices, with debates on whether models inherently amplify or merely inherit these biases.
Activity Over Time
Top Contributors
Keywords
Sample Comments
How are you managing human bias in this model?
Am I thinking to simple that it's obvious? Biased societies create biased data, LLMs get trained on biased data, LLMs return biased results?
LLMs are not magic, they inherit whatever bias is in its training data.
Can anyone ELI5 what the source of bias is outside of the biased training data?
Its more like " There are difficulties on collecting data and historical reasons that leads to bias in the trained model" Other side " no you are a racist and responsible for bias due for being a white male"
I don't even think he said "model's don't cause bias," he just said "ML systems are biased when data is biased."
This is awesome! What kind of bias would we expect to see in the training set?
So the bias is left to the AI's training set?
It's crazy the model can learn implicit biases even when you don't supply that data!
I'm not so sure. seems to be human bias on top of data bias. great for context https://www.youtube.com/watch?v=gdCJYsKlX_Y