As of this month, as I track many communities and disciplines and global challenges (like the 7 on my list attached)... I am astonished that Elon Musk might be our best hope after all, right now, unless things change.
None of us is perfect. None of us knows everything we NEED to know. But Musk has spoken up for critical goals needed to understand and survive the first great challenge on my list, the challenge of surviving a complex of risks coming from internet/AGI/ IOT. In particular, see:
This news story exemplifies how badly the usual policy people, investors and press misunderstand AGI. The recent security council session on AI was much closer to the reality of the new technology (because it had inputs from the world's strongest defense technologies,), but most "expert opinions" are very delusional and dangerous. Many of these opinions came from the experts who told us that deep learning could never do anything important. (See
werbos.com/mind.htm for a link to a video by Sergey Brin telling us how he was stunned, and how he changed the game back then.)
Musk's new plans simply reflect a few realities that our policy experts and dangerous delusions ignore. Above all, they do justice to the importance of the basic unlearned utility functions U which dominate all true AGI.
It is those choices of U, not childhood experience or regulation by stupid inherently ignorant bipeds, which will decide whether we all live or die, starting well before the end of this century.
When humans propose to solve the risks by demanding explainable A,I, they prove they do not understand the immense science already done about explainable AI. Should we just outlaw all AU which understand and account for general relativity, even if those AI are unable to make average humans understand that subject? If 90% of nations outlaw such AGI, but ten percent use them, which will manage the planet in the end?
Musk has not addressed all the life or death details. But at least he is asking about U and about basic realities, without which we will have no hope at all.
Of course, he also asked about reusable access to space and quantum method to control fleets of vehicles. If he ends up being as well motivated but insufficient on details, he will screw up the final bottom line as badly here as there. If so, we all die. Probably so. But there is some hope, which is better than nothing. I will never forget the woman I met who helped solve Tesla's problems with batteries in the first iteration of that effort. There is some basis for hope. Let us, like her, try to make the hopes become real.
My worries about the alternatives are based on lots of deep details on IBM, Google and Microsoft, and THEIR pathways on AGI and quantum technology.