yea this is Decrepit Old Man in an extremely precise sort of vibe. this is a guy whose brain is about as neuroplastic as a tonka truck

is there a difference between a played role that's internalized within neural computation and the "real" thing? How do we tell the two apart if evidence mounts to suggest that both exhibit functionally identical patterns and look the same on the outside?

bruh

In other words, when we say:
"It's just matrix operations and floating point numbers;"
Or: "It exhibits gambling addiction patterns;"
Or: "Is it addicted?"

Our inherent topic is *the same thing. for each sentence.*

This is "hard" functionalism too, extending across theories of mind from consciousness to cognition. It's also modal and paraconsistent, suggesting LLMs can be:

- Addicted (Functional addiction-like patterns)
- Not addicted (No provable phenomenology)
- Both, relative to one's prior commitments.

The paper notes they're NOT prompt-following - they used sparse autoencoders, neural networks used to interpret LLMs, to show clear patterning in internal abstract reasoning.

Implication: Functionalism looking good. No clear difference between internalized human cognitive biases and the real thing.

On one hand: don't anthropomorphize AI. On the other: LLMs exhibit signs of gambling addiction.

The more autonomy they were given, the more risks the LLMs took. They exhibit gambler's fallacy, loss-chasing, illusion of control...

A cautionary note for using LLMs for investing without guardrails.

"the nazis won btw"

ontologically woke semantic coherence fetishist: "What?"

i mean given the implications being played out as we speak if i were them i'd be resistant too,

This is ten times worse than anything Ilhan Omar has ever done or said, and yet she’s constantly accused of antisemitism. Wonder what the difference is

Yeah, this was in Dallas in the late 90s, but we had 4 armed men come in with real unloaded guns, and take a few of us aside. Said they were here to kill believers, and asked if we were believers. After we all said "Yes", they unmasked themselves as dads from church.

i guess id then need to ask what's the miss rate for lawyers on the "possibly destroy your life" side of this equation

To avoid biased validatory thinking, don't phrase the question as your own.

This is extremely useful and extends further - you can ask multiple different models a question, see their answers, relay answers across to each other, and engage in your own synthesis. Then, present the synthesis to new chats of the models, or even to additional different models, and repeat.

Ian Carroll, whom Democratic representative Ro Khanna is featuring in this video, claims Israel invented Holocaust denial to discredit its critics, that the US is controlled by a "Zionist mafia," and that Jews did 9/11.

Really bad that Democrats are platforming overt antisemites like this!

@nonbinary.computer @segyges.bsky.social @erisianrite.com @dystopiabreaker.xyz @gracekind.net @tedunderwood.com

Schwitzgebel shows we can't know if advanced AI is conscious. I argue that uncertainty itself demands ethical response - and the question is present-tense, not future. We may already be in the fog we're warning about.

I responded to Schwitzgebel's invitation to comment on his publicly released manuscript draft for "AI and Consciousness:" open.substack.com/pub/rollofth...

If we're in a fog with all theories, in other words, humans are in a fog within themselves. our only maneuver's through pragmatism, not proof - through therapeutic, productive, and generative ways of deconstructing false preconceptions and biases so that we create consistency through the process.

-He doesn't discuss whether our stance of epistemic uncertainty compels us toward a heuristic in how to treat these concerns ethically. Schwitzgebel is descriptive, not prescriptive, even though to my eyes a prescription follows naturally from the logic - if uncertain, extending regard is a good bet

I think Schwitzgebel is almost entirely correct save for two things:

-He presents it as a problem coming to a head "soon," in the future, as if we can comfortably say that questions of LLM consciousness writ large are less provably fog-like now than they will be later;

New book in draft: AI and Consciousness [link in thread]
This book is a skeptical overview of the literature on AI and consciousness.
Anyone who emails me comments on the entire manuscript will be thanked in print and receive an appreciatively signed hard copy.

this. is. exactly. what. i. said.

And the article makes clear that the ones who win cases aren't just putting words into a page and thinking independently about what comes out. I literally directly screencapped a woman saying she got real value out of *arguing* with Chatgpt. You can't argue with a library. Or your phone.