I downvoted this post, even though I appreciate the effort you want to put into explaining the risks and timelines of AGI to outsiders of CS. This is for several reasons:
I generally oppose “persuading person X to believe in Y”, unless that person is a relevant decision-maker and the belief is something that we understand very well why it is true.
The aggregated community/expert prediction shouldn’t be taken as anything other than it is, and is definitely not something we should be sure about ourselves. In fact, views regarding exactly this question vary widely from person to person.
Persuading people that some future event is p% likely seems weird to me, in this case. I’m not sure exactly why, maybe at least something like “without knowing the particular person’s priors we can never send them information to convince them of any particular probability”, but I think it’s articulateable without bayespeak.
Actually, this is not just for persuading others, it’s actually persuading myself. I a CS outsider, and I really don’t understand why many people are confident that AGI will be created someday.
2.Of course predictions may not be accurate, and it’s a personal view. But I think there must be some reasons why you predict AGI is 50% in 2040, not 10%.
Those are great questions, sorry if I misinterpreted your original intent.
This is a technically complicated question, but I think Holden Karnofsky tackles those questions rather well in his blog. Say, you could start from this post and read back to the links that interest you.
I downvoted this post, even though I appreciate the effort you want to put into explaining the risks and timelines of AGI to outsiders of CS. This is for several reasons:
I generally oppose “persuading person X to believe in Y”, unless that person is a relevant decision-maker and the belief is something that we understand very well why it is true.
The aggregated community/expert prediction shouldn’t be taken as anything other than it is, and is definitely not something we should be sure about ourselves. In fact, views regarding exactly this question vary widely from person to person.
Persuading people that some future event is p% likely seems weird to me, in this case. I’m not sure exactly why, maybe at least something like “without knowing the particular person’s priors we can never send them information to convince them of any particular probability”, but I think it’s articulateable without bayespeak.
Thanks for your response and patience very much.
Actually, this is not just for persuading others, it’s actually persuading myself. I a CS outsider, and I really don’t understand why many people are confident that AGI will be created someday. 2.Of course predictions may not be accurate, and it’s a personal view. But I think there must be some reasons why you predict AGI is 50% in 2040, not 10%.
Those are great questions, sorry if I misinterpreted your original intent.
This is a technically complicated question, but I think Holden Karnofsky tackles those questions rather well in his blog. Say, you could start from this post and read back to the links that interest you.
Thanks. This helps really lot.