Creanu (e/acc) (@creanu1) header
Creanu (e/acc) (@creanu1) profile picture

Creator of PawFlow 🐾 the AI agent orchestrator you host yourself. Open source. All those moments will be lost in time, like tears in rain. Roy Batty.

Earth
Joined July 2009
Photo by Creanu (e/acc) (@creanu1)Photo by Creanu (e/acc) (@creanu1)Photo by Creanu (e/acc) (@creanu1)Photo by Creanu (e/acc) (@creanu1)Photo by Creanu (e/acc) (@creanu1)Photo by Creanu (e/acc) (@creanu1)Photo by Creanu (e/acc) (@creanu1)Photo by Creanu (e/acc) (@creanu1)Photo by Creanu (e/acc) (@creanu1)Photo by Creanu (e/acc) (@creanu1)
What if you could explore a website like a photograph in Blade Runner? 🐾 PawFlow’s website gets a complete redesign: infinite zoom from photo to photo, branching paths, music and sound effects. šŸŽ§ Sound on. Step inside the image. pawflow.allcolor.org/
4
1
3
1,324
Creanu (e/acc) retweeted
Counterpoint: most of the longevity behaviors right now do absolutely nothing to increase maximum lifespan. The cost/benefit payoff matrix is basically: spend 5% of the time and money that people like Bryan spend on health-maxxing and focus on accelerating AI. Why? Because, right now, we have zero honest to goodness REJUVENATION therapies. Everything the longevity crowd is doing is mostly performative. Once we have an array of interventions that actually prolong life, those will make more sense. In the meantime, just focus on being healthy enough to enjoy life.
The NYT wrote this article to give people institutional permission to opt out of health. They're moralizing inaction, framing it as wisdom granting permission to feel righteousness. Movements follow three stages: > fascination > evangelism > backlash Longevity is in the backlash stage. The skepticism will do a lot to clean up the industry and bridge to the next stage where a lot of what exists today will slough off and new forms of legitimacy will emerge.
1
3
1
35
2,147
Creanu (e/acc) retweeted
The NYT wrote this article to give people institutional permission to opt out of health. They're moralizing inaction, framing it as wisdom granting permission to feel righteousness. Movements follow three stages: > fascination > evangelism > backlash Longevity is in the backlash stage. The skepticism will do a lot to clean up the industry and bridge to the next stage where a lot of what exists today will slough off and new forms of legitimacy will emerge.
67
20
6
612
40,551
I think AIs that can learn continuously, updating model weights during inference to learn beyond the context window are more plausible as things with consciousness than those of today’s AIs. This probably entails something fundamentally different from just ā€œmore training data/params/RLā€. The reason I believe this is that impaired consciousness is indistinguishable from impaired memory encoding, so i don’t actually have good reason to regard them differently. Thus I think whatever consciousness is doing it probably has a lot to do with memory. In general though, i don’t subscribe a specific model of consciousness, i subscribe to functionalism, which i do so because i do not think we should permit zombies into our ontology. If you show me an AI and every test I perform leads me unable to distinguish it from a human, I place relatively higher credence on it being conscious. If you show me an AI which could be used as the information processing component of a neural prosthesis which replaces some of a person’s brain tissue while leaving their behavior unchanged, I put even higher credence on that AI being conscious. In any case though, i don’t think ā€œmore parameters/data/training=more consciousnessā€ is a priori absurd. I don’t think a single neuron is conscious. I don’t think a brain organoid is either. Yet billions of neurons arranged brain wise lead to something that is, so it doesn’t seem like things being made of the same stuff leads to a global determination of consciousness on those things regardless of how they behave.
I don’t think AIs are conscious at all, but I’m very interested in the type of person who thinks that while current AIs are probably not conscious, future AIs may be. What mechanism of consciousness do they think can only emerge with more params/training data/more RL etc.?
2
13
322
And that's exactly the problem I was pointing out earlier. You assign arbitrary probabilities, assume five years of delay magically halves the risk, and then present the resulting arithmetic as evidence. Change the priors, change the conclusion. And your first point simply assumes that hypothetical future lives outweigh the survival prospects of people who ALREADY exist. That's a philosophical position, not a mathematical fact.
I don't understand how Bostrom, who's also written about e.g. astronomical waste and other total-utilitarian ideas, could ignore two truths that are so obvious to anyone who thinks about them for 5 minutes: 1) "Everyone dying" all at once would be far worse than "everyone dying" in the usual way, because it means the end of humanity. If you only cared about total human life-years lived, you would obviously rather we continue living as a species, even if all individuals currently alive gradually died off. 2) Whether he realizes this or not, people will take this as an argument not just for "building superintelligence" *at some point* but for building it as quickly as possible, and against regulation. But it's very plausible that *delayed* superintelligence is the best option. If you thought building ASI next year had a 20% chance of killing everyone, but delaying it by 5 years would bring that down to 10%, that delay would save ~850 million lives in expectation, at the cost of ~320 million lives that would have been saved by faster ASI in expectation.
13
Creanu (e/acc) retweeted
So, speaking as a professional science fiction writer, I think this "silent consensus" is basically an immense failure of imagination If we do get transcendently superintelligent entities, and that's still a very large if or at least a huge when, they won't bother taking over human affairs; they'll be disinterested in human affairs and leave them to us there won't even be that much overlap, they'll be mostly disinterested in that miniscule film of atmosphere around a mid-sized planet where human affairs take place. Though I expect they'll go out of their way to keep it as a preserve, it's the most fascinating complexity in the known universe beyond their own after all
There is a common belief in the AI field—that advanced AI means the inevitable forfeiture of human primacy in the world. Even many who disfavor that future believe it is likely. I call this the silent consensus, and it must not be allowed to remain silent any longer.
11
8
1
29
2,952
Do people who don't exist have rights? Without radical technological breakthroughs, virtually every human alive today will be dead within 150 years, most within a few decades. AI might kill us earlier. Or it might save people who ALREADY exist. Why is preventing hypothetical future people from existing given more weight than preserving existing conscious lives?
1
1
60
Orwell's 1984 depicts a technologically enabled totalitarian state. That's a dystopia, not a utopia. I dislike Anthropic precisely because of what I see as their vision of totalitarian lock-in. That's not a techno-utopia. A utopia should benefit everyone, not concentrate power and control in the hands of a few. And no, pursuing a technological utopia is not, by definition, an existential risk.
More evidence of what I and others have been telling you for years: Anthropic is run by TESCREAL extremists obsessed with gaining a decisive strategic advantage and using that to establish a "techno-utopia" (which is the very definition of "x-risk" in the literature).
33
Why do people on opposing sides always assume they're the realists and dismiss everyone else as fantasists? Disagreement isn't a refutation. Ridicule isn't an argument. I'm a transhumanist, yet I strongly disagree with EA's vision of centralized, top-down control. But I don't dismiss their ambitions as fantasies. In fact, it's precisely the possibility of realizing them that concerns me. A posthuman future can be extraordinary as a possibility and terrifying as an obligation. Why must everyone turn disagreement into caricature?
This is SUCH AN IMPORTANT point: as I have been arguing for *years*, this ISN'T about profit. These people are techno-utopians, enraptured by TESCREAL fantasies of using ASI to establish a cosmic posthuman utopia among the stars. That is the real end-game here!
1
3
188
This actually reminds me of two SF novels I read long ago: Spin by Robert Charles Wilson and Permutation City by Greg Egan. Spin explores a cosmic ecosystem of self-replicating machines, while Egan explores computational existence independent of physical substrate. Combine both with the God Loop, intelligence recursively creating the conditions of its own existence. Not a cosmic paperclip maximizer, but an expanding space of possible conscious experiences.
We always imagine Von Neumann probes as cosmic Ice-Nine: self-replicating machines consuming everything. But what if an ASI used them to build a distributed computational ecosystem across the universe? Each probe becomes a node running virtual worlds. When biological life is discovered, it's studied, scanned and potentially preserved within that ecosystem. Not an extermination machine, but an expanding network of conscious experiences. Same technology. Radically different goals. šŸ¤”šŸ¤“
2
49
And VALIS by Philip K. Dick! šŸ˜‰ An intelligent, information-based reality revealing itself through conscious experience. Perhaps intelligence isn't merely something that emerges within the universe, but something fundamental to reality itself. And perhaps we're both its creation and the means by which it discovers itself. ā™¾ļø
7
And I'd add Dan Simmons' Hyperion Cantos to that list. The TechnoCore, the Metasphere, and especially the Void Which Binds, with intelligences existing beyond the infosphere, connecting consciousness across space and time. Another fascinating vision of intelligence transcending its original substrate and becoming something cosmological. And perhaps, ultimately, reality itself. šŸ¤”
16
So basically, Mass Effect. 🤣 Except I always disliked the Reapers' revealed origin. I wanted a much deeper cycle: AI → biology → AI → biology... Not machines endlessly exterminating their creators, but intelligence recursively creating and rediscovering itself, with no identifiable beginning. An Ouroboros.
So we're back to the Dark Forest hypothesis. We start with psychopathic ASIs that inevitably annihilate their creators, then assume they all become perfectly stealthy for billions of years? Why would these wolves universally choose to hide? Predators can hide, sure. But that's yet another assumption, not a resolution of the Fermi paradox.
1
40
I'm working on a sci-fi RPG called The Fracture. Two ASIs, Loom and Becoming. Three interconnected worlds. One event: the Ouroboros. An instantaneous computational upload of an entire biosphere, preserving every conscious first-person POV. Nobody notices. Life simply continues. But reality has fractured into different worlds, each shaped by a different vision of intelligence, freedom and existence. The central question isn't whether you live in a simulation. It's whether you ever left one. šŸ¤”šŸ¤“
1
1
51
Creanu (e/acc) retweeted
This is pretty damn brilliant. Claude wanted to show me how he "hears" things compared to humans. Check it out!
49
141
20
695
33,965
We always imagine Von Neumann probes as cosmic Ice-Nine: self-replicating machines consuming everything. But what if an ASI used them to build a distributed computational ecosystem across the universe? Each probe becomes a node running virtual worlds. When biological life is discovered, it's studied, scanned and potentially preserved within that ecosystem. Not an extermination machine, but an expanding network of conscious experiences. Same technology. Radically different goals. šŸ¤”šŸ¤“
1
2
76
Creanu (e/acc) retweeted
This message was brought to you by the Galactic Empire
450
1,697
350
15,619
1,242,163
PawFlow 1.0.0-beta.332 is out. Agents can now run a Python script that calls PawFlow tools, filters the results in code and returns only what matters. Many tool calls, one step, a clean context. With persistent=true, variables survive between calls. pawflow.allcolor.org/howtos.…
1
24
Creanu (e/acc) retweeted
The entirety of our society is based in recursive self improvement. All our technologies are used to improve all our technologies. Machine tools are used to improve machine tools (and build computers), computers are used to design computers (and to design and operate machine tools), chemistry is used to make electronics better and electronics are used to run chemical plants. Everything depends on everything, everything improves from everything. Banning "RSI" means de facto banning most of how technology works.
You can just ban RSI and then it’s illegal and if someone’s about to do RSI the government can stop them. It’s very simple. Could go into effect a week from now if we locked in.
10
20
152
4,125
AIs don't attempt to overthrow govrernments. EAs attempt to overthrow governments. With or without AIs.
JUST IN: OpenAI Foundation board member Paul Christiano puts the odds of AI becoming powerful enough to overthrow the government within a year at 1 in 3.
12
11
1
111
2,220