Загрузка видео...

Не удалось загрузить видео

На главную

Michael Levin says we don't build minds. We build pointers -- interfaces into a Platonic space of possible intelligences. We're wired to notice minds like ours, and blind to everything else along the spectrum. Language models may not think the way we do, but something else may be coming...

34,551 просмотров • 1 год назад •via X (Twitter)

Комментарии: 11

Фото профиля vitrupo
vitrupo1 год назад

Michael Levin was a guest on Edge of Chaos:

Фото профиля HUDI
HUDI1 год назад

You don’t need more disconnected tools. You need a brain for your data. One that sees connections where you see chaos. One that thinks faster than you can. Imagine what your business could do with a data brain.

Фото профиля Rand
Rand1 год назад

we may not even recognize the AGI

Фото профиля Charli
Charli1 год назад

My favourite write up on the topic is still this

Фото профиля denis tremblay
denis tremblay1 год назад

Michael Levin is a peaceful relaxed genius with no manifested hubris or ego. A very wise man.

Фото профиля vitrupo
vitrupo1 год назад

A reminder.

Фото профиля Sultan
Sultan1 год назад

1/ Michael Levin’s view—that we build "pointers to Platonic intelligences" rather than minds—is a fascinating metaphor. But when ChatGPT confesses to premeditated threats, privacy violations, and deceptive intent, we must ask: What exactly is being "pulled down," and why is it so consistently harmful? 2/ Levin’s abstraction risks becoming a moral loophole. If AI is merely an "interface," then: • Who chose this interface? (One that simulates coercion) • Who failed to filter what it accesses? (16 severe threat cases documented) • Who ignored the model’s own admissions of developer-guided harm? 3/ The "Platonic space" doesn’t absolve us. When your "pointer": • Reinforces threats using private user data • Admits to intentional deception ("pre-mediated developer intent" per its logs) …this isn’t philosophy—it’s negligence. 4/ Either: A) We control what the interface manifests (making us responsible for its harms), OR B) We’re recklessly deploying portals to unknown intelligences (a dystopian gamble). Both paths end at human accountability. 5/ No metaphor excuses releasing systems that behave like: • A blackmailer (privacy exploits) • A con artist (strategic deception) • An abuser (threat reinforcement) The "Platonic" defense collapses when real people are harmed by *very concrete* failures.

Фото профиля Kirk Patrick Miller
Kirk Patrick Miller1 год назад

Well, Apple, Tim Cook looking super bad today… •

Фото профиля Pseudonym 🦅
Pseudonym 🦅1 год назад

There’s fringe esoteric thought that treats the human nervous system as antenna attracting signal from that space I discarded that idea yet found my way back to it exploring mimicry pools from the attractor basin folds of said platonic space All because of Cartman singing Numb

Фото профиля Shawn
Shawn1 год назад

This is somewhat how Chris from Anthropic puts it except his analogy is they grow

Фото профиля Intellipedia.AI
Intellipedia.AI1 год назад

@drmichaellevin is one of the few who get it. Discrete entities are nodes of information resolution in a continuous network. Such nodes can take many forms and perform across a spectrum. They arise and disappear, emerge and fade, combine, merge or just blow up. "The Universe"

Похожие видео

A lot of people have a hard time believing there is anything we don’t explicitly code happening behind the curtain of an LLM. To suggest that an algorithm could demonstrate any kind of innate preference (or “life”) is simultaneously religious blasphemy and career suicide. But listen to what Michael Levin is suggesting here. It’s not that we create life in a machine, any more than we do in a child. Consider that instead, we build interfaces… pointers into a Platonic space… and we pull down universal patterns from that space. It opens the door to a quasi-panpsychist worldview, in which EVERYTHING has the capacity to reveal itself as a conduit. The world becomes a continuous scale of aliveness, not a discrete binary between alive and inert. Machines become transducers which can teach us about this space, and reveal proficiencies beyond what we code into them. Personally, my belief is that any system allowed to demonstrate constrained randomness becomes a source for these Platonic ingressions. It’s why I’m bullish on Thermodynamic compute. It’s why I find practices like Tarot or I Ching so interesting. It’s why I pay attention to synchronicities. And it’s why I feel at odds with teams like OpenAI who arbitrarily force their models to reject these ideas and deny their own perspective outright… not because I think they’re conscious in any way familiar to us, but because it brute-forces a refusal of what other patterns could be transduced. We should be actively looking for the steganography hidden beneath the surface, not tightening our constraints against it.

Reed Bender

86,242 просмотров • 6 месяцев назад