Mark Neyer

Friendly yet highly disagreeable. Wildly open minded yet deeply opinionated. Ex Google, Facebook, &c &c.

Memoir out soon:

[Mark Neyer @apxhard]

"Good" is a vector pointing from t=0 to the distant future. You evolved drives and instincts to approximate it. You can't measure it b/c you don't have access to the future. If you don't think it's real, you're going to get yourself killed, either as an individual or group.

[Mark Neyer @apxhard]

@Grummz AI does a great job pointing the finger back at us, and saying “ you have no way to prove you are conscious either.” As if the shared philosophy of the west is incapable of recognizing the reality that we are conscious.

[Mark Neyer @apxhard]

I think the right conclusion is that language is inherently value-laden. It was obvious beforehand that language contains all possible things you could say. The question is whether what you said was worth reading.

What wasn’t obvious is that just training on all this text seems to have trained the model on a decent approximation of what humans actually value. That’s how they’re able to actually answer questions, they need to be able to give the most valuable answer, and they seem to do a decent job of it.

[Mark Neyer @apxhard]

@t_blom This is the problem I’ve been working on. Prototype here.

https://t.co/akHIXRNZeK

[Mark Neyer @apxhard]

@QiaochuYuan This is evidence against AI going FOOM and evidence that we should expect a bunch of different agents who are competing and cooperating to earn human trust

[Mark Neyer @apxhard]

@justalexoki You’re not supposed to BE him, you’re supposed to FOLLOW him. So the question isn’t, “what would Jesus do,” but rather, “Jesus, what are you calling me to do?” Then sit in silence while the question echoes back into the present .

[Mark Neyer @apxhard]

@VraserX Yup, I argued the ai safety movement increases AI risk:

https://t.co/CJ9brGwtJc

[Mark Neyer @apxhard]

@slatestarcodex @HumanHarlan It’s simple: you don’t know what configuration it will be, but you can assume it’ll be randomized since that’s what huge energy transfers tend do. Since it’ll be random, you can safely infer everyone is likely to die.

[Mark Neyer @apxhard]

@minordissent I agree, but think shaming adults for genuinely not knowing a better way is the same failure mode as spanking children.

[Mark Neyer @apxhard]

Some people are claiming that hammer time lasts forever.

This is misleading, experts say, because hammers don’t usually come with clocks in them.

[Mark Neyer @apxhard]

@mmjukic @pmarca Because scale wins in combat. An army of 10,000 slaves Beats 100 IQ-130 high trust persons. Everything in the west has been sacrificed to scale, whenever it got in the way of scaling.

[Mark Neyer @apxhard]

computational complexity is real, though: some problems are practically unsolvable as you scale them up. Others - like the halting problem - are unsolvable as a matter of fact.

There is also is chaos, which means "a system that might be predictable if you knew everything about it, but tiny amounts of ignorance compound over time." You can't predict where the moon, earth and sun will be ten million years from now because even simulating the dynamics of those 3 bodies can't be done perfectly over that time frame.

[Mark Neyer @apxhard]

@JayisDreeming @peterrhague What makes you think "we all want the same things?"

Many people want to 'win' or 'be the best', and even if they share that desire, it's inherely rivalrous.

[Mark Neyer @apxhard]

Why AI companies will build multiplayer games.

https://t.co/eqQsnEx7iy

[Mark Neyer @apxhard]

nope, i think there's an optimal way of being, and it involves most focus being local, and family being sacred, and a diversity of perspectives on how to scale, each one tied to a specific place in the world, which they protect. If that's the optimal way of being, "all of us being one" is an attempt to undermine optimality for a naive fantasy.

[Mark Neyer @apxhard]

@peterrhague The who notion of a flag originally came about to help soldiers figure out where their side was. The largest tyrannies possible are going to claim some notion of 'unity' or 'we are all in this together' because that's the most scalable meme you can use to coordinate violence.

[Mark Neyer @apxhard]

@eigenrobot Modern state liberalism really is just Protestantism taken to its logical conclusion. Once you reject tradition in favor of scripture, why not reject the bible in favor of 'evidence' and 'science'? Then, replace Jesus with the state, and you're off to the races.

[Mark Neyer @apxhard]

@eigenrobot I think it will evolve. The internet is like humanity's brain, yes, but so far all we have is the amygdala: just a bunch of images and emotion. I've been working on a prototype with the goal of building something like a shared neocortex.

https://t.co/akHIXRNZeK

[Mark Neyer @apxhard]

Why AI Won't Kill Us All: What the Orthogonality Thesis Misses

https://t.co/3tLtPpuV3Y

[Mark Neyer @apxhard]

@TheUnderSub @TRHLofficial he really wants that 5 star rating

[Mark Neyer @apxhard]

@antoniogm it's amazing how AI is bringing back text/cli interfaces. This is awesome.