An AI is built to make predictions and recommendations for outcomes (investing, hiring, where to move, financial policy, etc.)
This AI is trained without bias, out of vast amounts of reliable data.
Is it possible for this AI to be racist?
Conversation
Replying to
This question requires a definition of racist for the poll results to make any sense.
2
Replying to
Since it would likely reflect the inputs of the real-world: yes. Also depends on how loose a definition of racist was used.
4
Replying to
The simple no/yes reply lacks context which is probably necessary, but an awesome question either way.
1
Replying to
I voted 'yes', but to be fair what I envision would not necessarily be 'racism' as much as 'unjust discrimination' against particular groups. It may not be race-based at all.
1
3
Replying to
I'd say no, given that the AI won't be inferring causes behind its projections. The results would be heuristic and surface-level. Its results might look racist at first glance to the non-technically-minded, though.
1
2
Replying to
Inherently racist? No.
The results will still be skewed along racial lines, but it'll be because our society is un-equal and the data will reflect the generational and educational disadvantages suffered by minorities.
1
8
Replying to
Yes, obviously, because it’s being trained on a data set produced by a world with a long history of racial inequality.
4
Replying to
Poll is meaningless without an agreed-upon definition of racism. Everyone seems to have their own.
1
4
Replying to
Yes, because racism is rarely a full, systematic, belief. It's complex as belief, it requires some prerequisites to be held as. Racism is lack of ethics imbued into recurrent patterns: lack of BOTH intuition AND human empathy on same time. Something we expect from AI.
1