grayelephant

[grayelephant @grayelephant2]

@RokoMijic @ForHumanityPod Interesting as a "snap your fingers" plan but how do you suppose we would reach the political conditions required to implement anything so precise as this?

[grayelephant @grayelephant2]

@deanwball The AI Debate needs to reckon with the thesis that some proportion of the population has a stable lack of imagination. Given any information, they will not extrapolate. No day will come when they will flinch at the vistas of possibility before them, or even perceive them

[grayelephant @grayelephant2]

@matthewktromp @AndrewCurran_ Uncanny

[grayelephant @grayelephant2]

@ForHumanityPod Here's what I'd say, John:
1.) a lot of people are angry about AI in general, but not x-risk as much
2.) Those who aren't mad about it don't seem to strongly believe that the tech leaders strongly believe
3.) A sense of disempowerment tempers anger, and most don't feel empowered

[grayelephant @grayelephant2]

@Bunagayafrost @echonought Jesus Christ this is how much we've normalized Hiroshima. Putting that aside, things didn't turn out well for the reasons the extremists believed they would, and one can find another case to take if you need one where the outcome was actually not good by intention or by chance

[grayelephant @grayelephant2]

@RokoMijic ...I don't believe the father of the basilisk is just now recognizing this tine of the ASI Risk fork. But yeah it disturbs me that there doesn't seem to be nearly as much effort going into the "Sam becomes emperor" problem. X-riskers seem largely unconcerned

[grayelephant @grayelephant2]

@Quinurum If you're actively trying to update your world model based on AI news it is getting overwhelming. I'm trying to figure out a new tech-assisted information processing strategy so I can take some pressure off

[grayelephant @grayelephant2]

@ilex_ulmus I can sort of see what you're getting at in real-life terms if I squint, but, to clarify, what would Holly!Harry have done in this situation?

[grayelephant @grayelephant2]

@daganshani1 @DFine2002 If the value of using your platform is support_gained - backlash_created per messaging attempt, I'm not sure Bill Gates is the ally you want. There are many, many folks who see Gates as the Mouth of Sauron. Still an interesting development of course

[grayelephant @grayelephant2]

@AndyMasley @GaryMarcus I think the hair to split here is that Dr. Marcus is skeptical that we're on track for near-term AGI/ASI and his catastrophes are all based on handing too much power to systems that are inherently prone to unreliability. All rooted in his tech-level opinions on deep learning

[grayelephant @grayelephant2]

@perrymetzger @Valen_T_N And yet you say you don't want to do it with @liron even though this is his thing? Honestly surprised to find you have intellectualized arguments. These however many months of following you you've seemed a rabid partisan; now I'm not sure what you are or why you post as you do

[grayelephant @grayelephant2]

@lumpenspace That's a punchy little evasion of the point I was making but it won't do. Do you recognize that your post conflated the ideas or not?

[grayelephant @grayelephant2]

@lwoodhouse Still seems like the biggest crux is whether the tech we have will significantly accelerate machine learning research itself. If not, Mr. Huang's world looks more reasonable. If so...well, the results could be dramatic (understatement). Right now this is a contested question.

[grayelephant @grayelephant2]

@bayeslord I think it's the shape of the situation that does that.
- Possibly near-infinite upside. Possibly near-infinite downside.
- Breaks most cultural ideas of what life will be like
- Extreme, fractal uncertainty
- try to ignore it, you know it will come get you eventually

[grayelephant @grayelephant2]

@FrankieIsLost ...why assume everyone will get one?

[grayelephant @grayelephant2]

@NathanpmYoung Consider: you need someone in your corner who gets excited to see the culture respond, not fatigued. Spin is the order of the day. The game has begun.

[grayelephant @grayelephant2]

@losslandscape Sounds reasonable to me but OTOH a big incident will plausibly crush their competition for a while so it's not clear what your point is. Bigger and bigger incidents seem likely to me.

[grayelephant @grayelephant2]

@the_yanco @davidslifka It might help against the last-ditch ASI interruptions. Plausibly the "kill switch" exists in the form of large explosions.

[grayelephant @grayelephant2]

@RafaRuizdeLira Coxon quitting was an opening for Team Safety to overplay its hand. I see Doomers getting sloppier as they gain popular traction.