(But first, a common misconception is incorrectly thinking that we’re worried about human extinction from current AIs, rather than future AIs. See my handy FAQ) www.lesswrong.com/posts/uxzDLD... 3/16 - ThreadSky

stevebyrnes.bsky.social • 72 days ago

(But first, a common misconception is incorrectly thinking that we’re worried about human extinction from current AIs, rather than future AIs. See my handy FAQ) https://www.lesswrong.com/posts/uxzDLD4WsiyrBjnPw/artificial-general-intelligence-an-extremely-brief-faq 3/16

Comments

stevebyrnes.bsky.social•72 days ago

Here’s a “demo” that it’s possible for a large active AGI development project to have a technical plan that’s claimed to create nice docile AGIs, but the plan would actually make callous sociopath AGIs, and fixing it is an unsolved problem: https://www.alignmentforum.org/posts/C5guLAx7ieQoowv3d/lecun-s-a-path-towards-autonomous-machine-intelligence-has-1 4/16

stevebyrnes.bsky.social•72 days ago

Here’s a “demo” that it’s possible for a large AGI development project to decide that even TRYING to make nice docile AGIs is ALREADY overkill, because AGIs will just automatically be nice, again for reasons that don’t stand up to scrutiny. https://www.lesswrong.com/posts/ixZLTmFfnKRbaStA5/book-review-a-thousand-brains-by-jeff-hawkins#Does_machine_intelligence_pose_any_risk_for_humanity_ 5/16

stevebyrnes.bsky.social•72 days ago

(And here’s a “demo” that at least one powerful tech company executive might be fine with AGI wiping out humanity anyway.) https://observer.com/2024/06/argument-elon-musk-larry-page-ai/ 6/16

stevebyrnes.bsky.social•72 days ago

Here’s a “demo” that if you give random people access to an AI, one of them might ask it to destroy humanity, just to see what would happen. Granted, I think this person had justified confidence that this particular AI would fail to destroy humanity… https://decrypt.co/126122/meet-chaos-gpt-ai-tool-destroy-humanity 7/16

stevebyrnes.bsky.social•72 days ago

…but here is a “demo” that it’s possible for people to do experiments that threaten the whole world, despite a long track record of direct evidence that this exact thing is a threat to the whole world, wildly out of proportion to its benefit, and that governments may even fund them. 8/16

stevebyrnes.bsky.social•72 days ago

Here is a “demo” that, given a tradeoff between AI transparency (English-language chain-of-thought) and AI capability (inscrutable chain-of-thought but the results are better), many people will choose the latter, and pat themselves on the back for a job well done. https://arxiv.org/abs/2412.06769 9/16

Posting Rules

Be respectful to others
No spam or self-promotion
Stay on topic
Follow Bluesky's terms of service

Comments

Posting Rules

Reply