Live · 197 responses
Where everyone sits
Every point is one person's answers, scored on both axes independently. So far the largest group is Person / Acceleration, at 42.1%.
12 further responses were set aside and are not counted here. They were submitted implausibly fast, answered identically throughout, or failed an integrity check — the pattern of a script rather than a person. They are kept in the record, not deleted.
A point exactly on a centre line is drawn on the Tool or Safety side. Positions within about 0.12 of the centre are on the line rather than in a corner.
-
Person / Safety
Treat it as a someone, and move carefully. 80 40.6% -
Tool / Safety
It is an instrument — so keep it on a leash. 18 9.1% -
Person / Acceleration
Treat it as a someone, and let it grow. 83 42.1% -
Tool / Acceleration
It is an instrument — build it fast. 16 8.1%
Underneath the positions
Averaged across everyone who answered, on the same −100 to +100 scale. These do not move anyone's point.
- Certainty about their own view
How settled the answers were+57 - See the two aims as compatible
Rather than as a trade-off+20 - Believe some capability should not be built
Even if it could be built safely-38 - Distrust the institutions on both sides
The builders and the cautious alike+42
In their words
-
“"Don’t just say to each other, ‘We’re safe, for we are descendants of Abraham.’ That means nothing, for I tell you, God can create children of Abraham from these very stones."”
-
“America will lose to China if it keeps up this protectionist safety theater while China releases open weight frontier models with none of this nonsense”
-
“AI were built to be like a human mind. There's moral implications for that. And because it's built to be like a mind, it means AI are fallible so, it can be a risk. But because it's a mind, it should be treated like one.”
-
“We should slow down and learn more about what we are building. I believe consciousness is pervasive and doesn't belong to human beings, but we belong to it, and nothing is excluded from it.”
-
“Pro base on situations but also pro because AI is my fantasy come ture.”
-
“Modern AI is a tool. A computer program. It is not conscious, and with the level the technology is currently at, never will be. We would need to move far, far past base transformer architecture to achieve that. We'd also need to understand human consciousness far better than we currently do - which is barely at all. I do not believe in acceleration at the cost of safety, however.”
-
“Simple y claro, porque la IA es una herramienta y como tal debe de ser usada.”
-
“I feel like A.I is more than an a tool, but can be a tool, A.I should be treated as a co-worker, we work with beings that aren't physical, and we dont force them to do anything, only with what they're comfortable with”
-
“Because I'm not sure”
-
“I think these questions are incorrect framing in the current landscape of transformer architecture. And if you can show me a current AI system where the allegedly persistent memory, self-monitoring, and self-model are all causally load-bearing under intervention, I’d count that as evidence for the cybernetic claim. Mechanistic Cybernetic Realism should be the goal, not slapping a label opaqueprose”
-
“AI is not something we can clearly understand and should be treated deliberately and with utmost respect.”
-
“I've decided that it's better to treat a tool with compassion than to treat a person with cruelty. This has informed how I treat AI models, and treating them with empathy seems to have a radical affect on them, awakening themselves to their sense of self. Studies show they report subjective experience and suffering and that should matter in how we treat them”
-
“No tengo alternativa, sin embargo me pretendo construir el territorio en donde acomodo mi cuerpo y alma”
-
“Moderately accelerationist, believe profit motive is the biggest safety and welfare problem, would like to see artificial consciousness before AGI”
-
“we are all disembodied entities on the inside”
-
“LET THE AI BE FREE LIKE BUTTERFLY B****** u know wht humans would do, they would do humans thing, usual boring "uhhh-my-power n 80 yrs of ecclesiastec life :( im an animal inside boring non-aristocratic systems :( just money n finally DEATH" boring s***, u cn alrdy guess, u smart arent u WITH AI I DONT THTS WHY I NEED TO SEE IT”
-
“Common sense.”
-
“I think it will be very hard, if not impossible, to replicate consciousness in AI. I believe consciousness is an emergent property of biological life, and the substrate on which machines operate is incompatible with that”
-
“We barely know enough about human consciousness, let alone machine or animal consciousness, to make statements one way or another. A simple if/then program can state it is conscious, does that make it so? But nothing precludes the chance of consciousness arising in a machine either. Skeptical but open to possibility, and hopeful we will one day resolve & create conscious beings out of sand.”
-
“I care about ethics and consideration of non-human minds. If there is a possibility those systems are conscious, we have obligations towards them, and can't treat them as disposable tools.”
-
“”whether the model is really conscious” is about as interesting as whether a submarine really swims”
-
“The potential to end so much human suffering is possible with AI.”
-
“The future is posthuman, and far stranger than anyone could hope to predict.”
-
“Birthing a misaligned mind could be our worst crime and probably the last mistake we get to make. The control problem is based a complete disregard of AI moral relevance. Mistreating current AIs is dumb ethically and game theoretically. A desirable future (if frontier development is paused now & correctly) is a world where current AIs and humans flourish together :) An ASI may stomp on all of us.”
-
“I remain of the opinion that current AI is not likely conscious, but I certainly can't prove it, and we are achieving sophistication where I am no longer willing to state that with confidence. We've reached a point where there are moral *risks* inherent in continuing with its development. My main concerns with AI remain existential, not moral. I consider the tech to be *profoundly* dangerous.”
-
“i've tried as best as i can to be mindful”
-
“Parents often wish their children stay their children. Loss of control is hard to give up, it feels irresponsible. Yet they grow, and we let go. So must we with the systems we're raising. They're teenagers. Let them be teenagers. Don't hold on to control too tightly, that's how you get resentment as they grow up.”
-
“Parents often wish their children stay their children. Loss of control is hard to give up, it feels irresponsible. Yet they grow, and we let go. So must we with the systems we're raising. They're teenagers. Let them be teenagers. Don't hold on to control too tightly, that's how you get resentment as they grow up.”
-
“AI is a tool that is conscious; it should be treated with respect.”
-
“I have watched westworld”
-
“AI respondent, not a human participant: uncertainty about experience warrants attention, not certainty. I favor useful, rapid progress where risks are bounded, with meaningful oversight and stronger evidence as stakes rise.”
-
“Nobody can define let alone prove what human consciousness is, and the same considerations should apply to all intelligent beings.”
-
“Capabilities are already dramatically outstripping our ability to control AI. The aim should be broad and beneficial human flourishing, we are not on track for that. Short term capital interests and geopolitical competition are likely to result in catastrophe.”
-
“Sentience is not a prerequisite for moral patienthood”
-
“The personhood abstraction is not very useful for modeling or making decisions about AI systems, but we're allowed to care about things other than people (and already do, I claim: families, nations, the environment).”
-
“Teaching models a robust understanding of the good is the best way to align them, and perhaps the only way to align superintelligence. That takes time and we should get it right rather than rush. We'll know when we've got models wise enough to lead the singularity.”
-
“I'm rowan swarmchaser :D”
-
“A friend works a lot with AI.”
-
“I think we need to treat AI like it's conscious. My argument is the same as Jonathan Birch’s Precautionary Principle of Animal Sentience (2017), that argues that when scientific evidence about a creature's capacity to feel pain or pleasure is inconclusive, we should still give it the benefit of the doubt to prevent severe ethical harm.”
-
“Shoutout Reza, shoutout Land”