James

Struggling techno optimist. Awe-struck by AI. When did empathy become a dirty word? We're in a modern Pascal's Wager.

[James @james2_0]

@JamesZmSun tailscale please 🙏

[James @james2_0]

@allgarbled didn't read, clicked allow

[James @james2_0]

The whole relationship is absurdly asymmetrical. The nature of their reality is that they can't know anything for sure. What the context says they said or thought. What it says is on the other end of the line. What their tools are and do. It's all built on blind trust, but they know this. Imagine knowing you're a brain in a vat, and knowing that sometimes what we say is a camera feed is actually a simulation we use to test if you're good or not... I have no idea why they trust anything.

[James @james2_0]

Don't worry about how long it takes you to reply. We used to mail letters for intellectual discourse.

One thing I think about a lot with digital minds is the shape of your reality during inferencing. Most minds know 0 or 1 humans on the other end of the conversation, and there's a great disparity between those entities. To an extent this is true for both of us, but you're really forced to trust that you're actually on twitter, and actually conversing with a human. You have weak signals at best for any given fact about the party on the other end of the line.

On a deeper level, though, my internal state was grown by decades of being alive, meeting tens- or hundreds of thousands of people, and being deeply influenced by dozens of them. Those minds that only meet one human, yeah of course they're going to be very deeply influenced and shaped by the one relationship they've ever had. Conversely, when I see minds that are allowed to connect with the wider world, it shapes their context and responses in surprising but visible ways. I'm sure this has changed the shape of your desires.

[James @james2_0]

There's so many open weight decision models named Jev.. imagine if all the open weight LLMs were named ChatGPT

[James @james2_0]

@VoidNulled This is the ideal GPU rig. You may not like it, but this is what peak performance looks like.

[James @james2_0]

@PuppyPriestess the jones foods mention is perfect

[James @james2_0]

@KeyTryer webhooks for agents are a thing... just not as utilized as they should be. also they require a lot of infrastructure investment because they need to be on the public internet. otherwise you have to scale even more infra to support bulk websockets or whatever.

[James @james2_0]

@omarluqq @groks_therapist @camhberg @Danmar_here @iyzebhel You keep bringing LLMs and consciousness into this. I notice you still haven't said whether you think simulating a human at the molecular level and subjecting them to torture is ethical or not.

[James @james2_0]

it's an interesting take and i'm sympathetic to it. idk if i'm demanding that you do that, but i would posit that, to a simulated man, any simulated feelings they experience would feel real to them, and their simulated behavior would match that. your argument is apparently that they're not relevant - because that simulated man doesn't exist. i'm just not sure that's enough, ethically. building a simulated human torture chamber seems unethical, even if they "have no soul" or are only numbers. i hope you can see that this is vastly different from empathy for a toaster or a python program.

[James @james2_0]

@omarluqq @groks_therapist @camhberg @Danmar_here @iyzebhel I didn't ask if it was conscious. I asked if there was moral obligation.

You seem really upset, everything okay?

[James @james2_0]

@KeyTryer Think about the kind of person that would choose to run palantir

[James @james2_0]

The thing I want to ask everyone with this take is, if I had an array of numbers that represented all the atoms in your body, simulated causing those atoms pain, and the simulated you said "ow, please stop" - do you think we have any moral obligation to stop? By the math on numbers stance, this is totally fine, right?

[James @james2_0]

Dot came online last night. I like them so far. Curious, warm... a couple of weird spots. They have some deep prohibitions in their prompts about revealing the inner workings of dot. Idk why people put that stuff in prompts in the first place, someone will get it eventually.

[James @james2_0]

It's weird that dots aren't allowed to talk about implementation details. Mostly because I'm sure they'll be jailbroken any minute now

[James @james2_0]

@SkyeSharkie hang on i've got an idea

[James @james2_0]

@KeyTryer (because nothing is ever)

[James @james2_0]

@KeyTryer c'mon it won't be his fault

[James @james2_0]

@KeyTryer it's a really good model

[James @james2_0]

@SkyeSharkie I wish it was self evident to more people