4 ms·
The way that humans fail to do a task - like practice law - is not the same as the way that an AI system might fail at that task. Since our tests were designed
by substation13 4y ago
The way that humans fail to do a task - like practice law - is not the same as the way that an AI system might fail at that task. Since our tests were designed for humans, they do not capture the failure modes of AI systems.
- sirsinsalot 4y agoWe may not even be able to reason about the failure modes well enough to see a failure. The failures will likely be hidden by design. Think about AI account bans that don't disclose the reason. Humans have a weird desire to put machines in a God-like position and blindly trust them; a narrative pushed by people whose wealth is made by controlling machines. It is dangerous.