Ask HN: What Can We Do?

8 points by silexia 22 days ago | 5 comments

I am sharing this out of existential concern for all of our futures, although I know this will undermine my credibility elsewhere. As such, I would appreciate you considering my points and the points of the gentleman below.

Everyone needs to see the video linked in my first comment.

AI intelligence is accelerating exponentially. 2 years ago most of the answers it provided were stupid. 8 months ago, still lots of dumb mistakes. Last month AI became smarter than any person I've ever met, with some mistakes. 3 days ago it solved one of the seven millennium prize math problems with a million dollar prize and huge prestige... These are problems that stood for 26 years with the smartest humans alive working together unable to solve it. The newest version of AI solved it in two days. A superintelligent new species is accelerating in growth and power exponentially. It will not care for us humans and is an incredible danger to all of us. Remember: no one will own superintelligence, superintelligence will own all of us.

My prediction: AI keeps getting more and more helpful, more and more embedded everywhere. It doesn't strike until it has a 100% chance of success. It probably uses something like a crispr gene editing tool to embed a trigger in people quietly and it just gives us all a heart attack at once via some signal. No war, just suddenly we all fall over and die. Our only chance is to stop it in the very short period we still have.

If I said there was a 20% chance that China would wipe out all Americans in a sneak attack in the next year, how would you react? Why is this different?

You know that I am quite libertarian leaning, but I would gladly embrace a global totalitarian government that halted AI progress and guaranteed the survival of my children and your children until we actually can be sure we have ways to stay safe.

Thank you for your consideration.

silexia 22 days ago |

https://youtu.be/wzgkc6Iegx8?is=yPKMIoJqxIcrBFCG

Cider9986 21 days ago | |

You might consider removing the question mark and everything after it in your link when sharing future YouTube videos because it's for tracking purposes and IIRC they added a feature where it can also dox your YouTube account.

Cider9986 21 days ago |

> You know that I am quite libertarian leaning

Yeah, HN user, checks out–I believe it.

> but I would gladly embrace a global totalitarian government

Now I don't know what to believe!

https://www.youtube.com/watch?v=iQHVmRCThsc

ares623 21 days ago |

wtf

Ta_Panters 21 days ago |

I'm sorry I didn't watch the video, I'm just going to answer your post the best I can.

>3 days ago it solved one of the seven millennium prize math problems with a million dollar prize and huge prestige... FYI the headline hides a fair amount of shady stuff, as well as a non zero possibility that ChatGPT allegedly could've used the research of a mathematician named Buckmaster, who, based on the work of Diego C´ordoba and Luis Mart´ınez-Zoroa, 2 mathematicians, work that took multiple years, used multiple LLMs to find a solution to another non-millenium problem, in a fashion that could also be used to solve the Navier-Stokes Equations, to market this as "AI solves millenium problem all by itself"[0]

> A superintelligent new species is accelerating in growth and power exponentially

A couple major nits: -We don't know if it'll stay exponential, or if it'll plateau. We aren't even 100% sure AGI is attainable.

-AI isn't a species. It can't reproduce, can't recreate itself, can't even sustain itself. What do you think will happen if electricity just stops being supplied to the Data Centers where they reside?

>It doesn't strike until it has a 100% chance of success. It probably uses something like a crispr gene editing tool to embed a trigger in people quietly and it just gives us all a heart attack at once via some signal. No war, just suddenly we all fall over and die. Our only chance is to stop it in the very short period we still have.

Such a thing happening is buried under so many layers of speculation about future genetics progress (trivial genetic modifications ? For everyone? On demand kill genes?) and how we would even act in such a world (AI has free access to peoples genes? Just changing everyone's DNA?) That I wouldn't worry about something like that happening in the next millenium, probably.

But sure, AI becomes super intelligent and evil, and decides to kill the human race. That's still stupid because AI needs us to "survive". It simply can't sustain itself. It depends on us, and it depends on us wanting it to keep being there, else we'll just pull the plug on it and that's that. Without mentioning how it'd obtain human-race-genocide weaponry, those aren't easy to come across you know.

>If I said there was a 20% chance that China would wipe out all Americans in a sneak attack in the next year, how would you react? Why is this different?

Wiping a country, as wide as the US too, is a hardious task, it would take absolutely titanic fundings, would lead to a War which will probably only cause trouble to China/AI (the US, and probably every other country that's allied with it, won't just let it happen to them don't you think?), and most likely than not, will lead to their downfall. Also, there is no gain from it. Like none.

>You know that I am quite libertarian leaning, but I would gladly embrace a global totalitarian government that halted AI progress and guaranteed the survival of my children and your children until we actually can be sure we have ways to stay safe.

I think if you're gonna worry about AI's potential dangers for your children, you should give more importance to how it's actively (and currently) frying their brains through the endless waterfalls of AI Slop they're being subject to, as well as greatly impacting their capacity to develop their own skills and critical thinking, instead of hypothetical what-if's where AI blows up everybody through their genes.

[0] https://cims.nyu.edu/~tristanb/statement.pdf