5 ms·
You could make a neural network that categorised those who are more likely to be criminals based on this data. Big if, but if it was accurate to say 95% would
by andy_ppp 8y ago
You could make a neural network that categorised those who are more likely to be criminals based on this data.
Big if, but if it was accurate to say 95% would it be moral?
Strange times in which we live.
- Anon84 8y agoWhy do you expect someone's voice to give clues as to wether or not s/he's a criminal? More likely than not, the model would simply learn to detect minority accents and further help institutional discrimination against minorities (which are already significantly more likely to be sentenced to prison, etc).
- roywiggins 8y agoI expect it would just categorize all younger men with African American accents and a gradeschool vocabulary as criminals. 37.1% of 20-to 34-year old black men without a high school diploma or equivalent are currently incarcerated. https://www.pewtrusts.org/~/media/legacy/uploadedfiles/pcs_assets/2010/collateralcosts1pdf.pdf https://www.pewtrusts.org/~/media/legacy/uploadedfiles/pcs_a...
- andy_ppp 8y agoWoah 37% is terrible, good point about the racism it would no doubt match against... My real question is as we get better AI systems and more accurate ones should we use them to make decisions that can affect people in profound ways. Seems that maybe this is not the right place to discuss something as open ended and off topic as this!
- thucydidesofusa 8y ago>Woah 37% is terrible what do you think the right number should be?