When the first @sanjayuvacha account lock happened, some people including the good lawyer himself thought it was an AI/algorithm error and I didn't open my mouth. He has been locked again due to some inanity.
Something I must stress, it is NEVER the AI's "fault". Hear me out.
Something I must stress, it is NEVER the AI's "fault". Hear me out.
Do you blame your stapler for racism? Your dining table for bigotry? The public discourse on AI and algorithmic bias is a caricature which needs to be pushed back by people who know the field.
The responsibility of the act of bigotry (and it is always an act) is always human.
The responsibility of the act of bigotry (and it is always an act) is always human.
It doesn't matter if an algorithm caused Hegde's account being locked (not that I believe it for a second), if there is no human moderation at that step @TwitterIndia is terribly stupid.
Don't let people and corporations slide off their responsibility on inanimate things.
Don't let people and corporations slide off their responsibility on inanimate things.
Let me be very clear here.
One of the larger reasons why I oppose surveillance by facial recognition tech, completely without quarters, is NOT bias. It is this erasure of responsibility.
The real danger is law enf saying it was not our fault, that machine told us to do blah.
One of the larger reasons why I oppose surveillance by facial recognition tech, completely without quarters, is NOT bias. It is this erasure of responsibility.
The real danger is law enf saying it was not our fault, that machine told us to do blah.
You can NOT remove bias from machine learning. And you can not remove error from something by design nondeterministic. And no you can't make it fair or accountable or transparent after a point either.
What you can do is not have it without adequate human supervision.
What you can do is not have it without adequate human supervision.
And in certain cases, where the very existence violates how lawful if not civilised societies operate, not have it. Facial recognition for surveillance should not exist. "automated" weaponry should not exist. No debates on this will be entertained.
I wrote this because some people were commenting on how this is the fault of techbros who are not educated in humanities and history and what not. No, it's not their fault. That's now how machine learning works.
And it's not your fault for assuming so I guess
And it's not your fault for assuming so I guess
*not
This may have happened because the human moderators (if any exist) who use the algorithmic suggestions are ill paid and tired and don't have clear guidelines etc. Or are plain dumb. That is the best case. You all know the worst case.
This may have happened because the human moderators (if any exist) who use the algorithmic suggestions are ill paid and tired and don't have clear guidelines etc. Or are plain dumb. That is the best case. You all know the worst case.
Okay there is so much bad public discourse on AI governance it's irritating me. Will one of the many journo type people who follow me or their newspapers like a critique of the "critiques" of AI which abound? Might as well write a series. My credentials are in the bio.
