I know that MIRI doesn’t aim to actually build a FAI and they mainly try to provide material for those that seriously try to do it in the future. At least this is my understanding, correct me if I’m wrong.
But has the work MIRI has done so far brought us closer to building a working FAI or any kind of AGI for that matter? Even a tiny bit?
From what I’ve seen in the last year, MIRI has sort-of backpedaled on the “actually building an AGI/FAI” goal, and pushed forward in their public declarations the “milder” goal of ensuring a positive impact of AGI when it finally gets created by someone.
Are you sure? The linked comment was posted in August of this year. I have seen them push the second goal more strongly recently, but I don’t think the first goal has been rescinded.
I know that MIRI doesn’t aim to actually build a FAI and they mainly try to provide material for those that seriously try to do it in the future. At least this is my understanding, correct me if I’m wrong.
But has the work MIRI has done so far brought us closer to building a working FAI or any kind of AGI for that matter? Even a tiny bit?
They do intend on actually building an FAI, but that doesn’t preclude them from also trying to provide material for others who might do it.
From what I’ve seen in the last year, MIRI has sort-of backpedaled on the “actually building an AGI/FAI” goal, and pushed forward in their public declarations the “milder” goal of ensuring a positive impact of AGI when it finally gets created by someone.
This seems consistent with the first sentence in their mission statement:
Are you sure? The linked comment was posted in August of this year. I have seen them push the second goal more strongly recently, but I don’t think the first goal has been rescinded.
It’s more of an impression of mine than an actual statement of theirs.