Paul Christiano, who is an OpenAI Foundation board member, just shared a pretty serious warning: he thinks "AI accidents" might become one of the most likely reasons they will die in the near future.
At a recent talk on AI ethics, he estimated there's a "something like a coin flip of an AI takeover" if AI is developed "along the most rapid trajectory," and about a 1-in-5 chance that major problems with AI could seriously disrupt everyday life.
Paul Christiano's OpenAI role heightens concern
Christiano's role at OpenAI makes his concerns hard to ignore.
His warnings echo what some online forums, like LessWrong, have been saying: if we don't get AI safety and rules right, things could go very wrong.