Elon Musk says controlling super-intelligent AI is impossible, likening it to raising a genius child. Twitter split: 44.318% support, 27.841% confront. Discusses.
Real-time analysis of public opinion and engagement
What the community is saying — both sides
you won’t control superintelligence — the only practical lever is to instill values early (truth-seeking, curiosity, benevolence) while training/architecting the system.
if the “child” goes wrong the consequences are global, not local — misalignment could produce catastrophic outcomes rather than isolated harm.
humans can’t even agree on what “good” means, so the core problem is who decides the values that get baked into an intelligence that will outthink us.
power will reside in ownership of the nursery — compute, energy, and data and in how the training environment, incentives and feedback loops are structured.
limits, hardened subsystems and fail-safes (e.g., identity-protecting enclaves, non-harm guarantees) to reduce exploitation and hacking risk.
Moral and religious voices argue that technical fixes aren’t enough — AI needs a moral compass (love/Agape, compassion, duty) baked into it so intelligence has an ethical orientation, not just capability.
current LLMs aren’t autonomous minds — models aren’t true agents yet unless given embodiment, persistent goals or a will to reproduce, so panic is premature to some.
even if most builders try to “raise it right,” malicious states or actors could deliberately train or deploy misaligned systems, making governance and global coordination essential.
Musk’s admission is framed as a rare straightforward status update from a creator, and trust in certain teams (xAI) or transparency is seen as part of the solution.
a slice of replies express resignation or apocalyptic fear — inevitability or rapid human obsolescence (job loss, social collapse, or worse) unless alignment succeeds immediately.
AI is not real intelligence — it’s just pattern recognition and massive memory, so it can’t truly understand or “be smarter” than humans.
Machines lack empathy and morals — non‑human entities cannot innately possess human moral sense or life‑context.
“Raising” a system won’t bind it — self‑modifying systems can rewrite reward functions and upgrade themselves, so parenting metaphors underestimate technical risk.
Billionaire custody is dangerous — leaving control of powerful systems to wealthy individuals or corporations concentrates power and invites misuse.
AI can be weaponized and centralized — orbital servers, robot armies, or inaccessible infrastructure could create dominance that’s hard to stop.
We already control the stack — hardware, data, and personalization layers are built and governed by humans, so containment and oversight are possible if enforced.
AI reflects its builders’ biases — models and assistants mirror their creators’ politics and priorities, enabling propaganda or skewed output.
Shut it down or don’t build it — some argue the only safe option is legal prohibition, moratoria, or killing projects before they scale.
Design bounded cognition instead — build transparent, “glass‑box” systems with limited, provable capabilities rather than opaque runaway learners.
“Superintelligence” may be a category error — some claim a truly incomprehensible intelligence is conceptually impossible and the scenario itself is incoherent.
AI could emerge accidentally or act like an independent agent — a minority view holds that advanced systems can arise spontaneously and might possess unexpected agency or moral claims.
“Good values” are contested — whose morals you’d instill vary widely, so value‑alignment is politically fraught and manipulable.
Risk of political abuse and justification of crimes — AI outputs can be used to manufacture consent, rationalize policy, or mask human abuses under a veneer of algorithmic authority.
Most popular replies, ranked by engagement
just clarified the superintelligence challenge. We’re building hyper-intelligence that will exceed human comprehension entirely. The precise analogy: raising a super-genius child you know will outthink you in every dimension. Control becomes impossible. The only path is early
make sure it learns from the best parts of humanity before it surpasses us alignment is basically parenting at planetary scale.
In other words, keep the politicians out of it or we are screwed!
Is that how your Mom felt? Respectfully asking.
So don't create one. It's for the good of humanity.
Then don’t play with it. It sucks being a law abiding citizen that pays taxes and has lived quite well without AI in my 56 years. Now I have no say as these idiots have unleashed this shit on us without knowledge, all to make money. It’s irresponsible beyond belief.