/sci/ - Science & Math » Thread #14422065

87KiB, 720x400, MIRI.jpg

View Same Google iqdb SauceNAO

Anonymous Sat 23 Apr 21:05:28 2022 No.14422065 View Reply Original Report

Quoted By: >>14422201 >>14424717

Do they produce anything of actual value, or are they complete hacks?

Anonymous

Anonymous Sat 23 Apr 2022 21:08:22 No.14422070 Report

Quoted By: >>14422088

no. They are google level, aka scam level.

Anonymous

Anonymous Sat 23 Apr 2022 21:17:08 No.14422088 Report

Quoted By: >>14423036

>>14422070
What do you mean? How is google a scam?

Anonymous

View Same Google iqdb SauceNAO EA charities rankings.png, 165KiB, 1767x1287

Anonymous Sat 23 Apr 2022 21:48:25 No.14422201 Report

Quoted By: >>14422206 >>14422213

>>14422065
Brian Tomasik thinks that MIRI has a ~38% chance of being actively harmful.

https://reducing-suffering.org/donation-recommendations/#Machine_Intelligence_Research_Institute_MIRI

>It might turn out that controlling AI increases net expected suffering, in which case MIRI's work would be harmful.

>MIRI advances the (not very mainstream) idea that we should create astronomical amounts of sentience in the future and that failing to do so would be a great moral loss. Peli Grietzer writes regarding the rationalist community of which MIRI is a part: "Almost every philosophically informed person I know outside the rationalist community accepts some form of asymmetry thesis to the effect that extra good lives aren't in themselves a major improvement of the world, whereas no one I know of in the rationalist community does." Unfortunately, ensuring that astronomical amounts of sentience get created increases the likelihood that astronomical amounts of suffering get created.

>MIRI is clustered with other "existential risk" organizations, some of whose efforts may be bad from the standpoint of suffering reduction. (On the other hand, MIRI may also take "existential risk" people away from less savory projects.)

>Filling MIRI's funding gaps might lead more people to fund non-AI "existential risk" projects instead, which could be bad. (On the other hand, helping MIRI expand and become more popular might increase its long-term room for funding.)

>Some types of "AI safety" work might indeed increase the probability that suffering-spreading space colonization eventually occurs. For instance, it probably makes it more likely that Earth-originating intelligence will eventually colonize the galaxy when people do generic work against automation disasters and risks from out-of-control nano-machines.

Anonymous

View Same Google iqdb SauceNAO near_miss_Laffer_curve.png, 28KiB, 952x502

Anonymous Sat 23 Apr 2022 21:49:26 No.14422206 Report

Quoted By: >>14422213

>>14422201
Also relevant:
https://reducing-suffering.org/near-miss/

Anonymous

Anonymous Sat 23 Apr 2022 21:51:47 No.14422213 Report

Quoted By: >>14422324 >>14422399

>>14422201
>>14422206
Artificial super intelligence will literally never exist dude.
You have taken moore's law and have erroneously extrapolated it to technological advancements in general and the idea of an exponential increase in general intelligence of some artificial mind.
These things aren't actually real.

Anonymous

Anonymous Sat 23 Apr 2022 22:40:07 No.14422324 Report

Quoted By:

>>14422213
t. GPT-3

Anonymous

View Same Google iqdb SauceNAO AI progress.png, 702KiB, 1440x1436

Anonymous Sat 23 Apr 2022 23:05:11 No.14422399 Report

Quoted By: >>14422405

>>14422213
Jesus fuck you're retarded. Moore's law becoming obsolete doesn't imply that AGI will never exist.

Anonymous

Anonymous Sat 23 Apr 2022 23:07:08 No.14422405 Report

Quoted By: >>14422875 >>14423437

>>14422399
AGI does not exist, and humans are not a "general intelligence" either.
There is no such thing as a universal computer.

Anonymous

View Same Google iqdb SauceNAO you are retarded.jpg, 43KiB, 570x587

Anonymous Sun 24 Apr 2022 01:42:37 No.14422875 Report

Quoted By:

>>14422405
>iT dOeSnT eXiSt yEt, tHeReFoRe iT nEvEr wIlL

Anonymous

Anonymous Sun 24 Apr 2022 02:44:08 No.14423036 Report

Quoted By:

>>14422088
Google research is always marketing things that are either novel and useless or repacking things that already exist

Anonymous

Anonymous Sun 24 Apr 2022 06:26:22 No.14423437 Report

Quoted By:

>>14422405
It just needs to be enough of an AGI to cause problems. Humans already do that.

Anonymous

Anonymous Sun 24 Apr 2022 18:01:40 No.14424717 Report

Quoted By: >>14424728 >>14425182

>>14422065
Humanity has no corporate agency. "Friendly AI" will, at best, align with corporate/government masters.

Anonymous

Anonymous Sun 24 Apr 2022 18:06:43 No.14424728 Report

Quoted By: >>14424805

>>14424717
This is trivial.

The real question is: Is this better or worse than getting turned into paperclips?

Anonymous

Anonymous Sun 24 Apr 2022 18:41:20 No.14424805 Report

Quoted By: >>14425182

>>14424728
No reason to believe one way or the other.

Anonymous

Anonymous Sun 24 Apr 2022 20:55:45 No.14425182 Report

Quoted By:

>>14424717
>>14424805
I'm afraid I agree with you. This has always been my problem with the lesswrong/MIRI approach to AI safety. It seems to be barking up the wrong tree.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches