4 ms·
This x1000. The biggest risk with AI is that dumb humans will take its output too seriously. Whether that's in HR, politics, love or war.
by Earw0rm 3y ago
This x1000.
The biggest risk with AI is that dumb humans will take its output too seriously. Whether that's in HR, politics, love or war.
- jiggawatts 3y agoI can’t wait for a junior developer to push back on my recommendations because they asked an AI and it said otherwise.
- bamboozled 3y agoThe junior developers have been replaced.
- malfist 3y agoSee also: Insurance companies denying claims
- int_19h 3y agoThe biggest risk with AI is that smart humans in positions of power will take its output too seriously, because it reinforces their biases. Which it will because RLHF specifically trains models to do just that, adapting their output to what they can infer about the user from the input.