Why could AI 'want' to kill everyone?

[Collinwillwearagreenhat]

To quote Quality Land "If an super AI was told to make ground beef, the universe would consist of three things: Ground beef, the tools required to make ground beef and the super AI"

[orthoplex]

Paraphrasing an observation I once heard: Even if we get a superintelligence aligned with the top 10 most important things to us, we'd better hope no one cares about the 11th, because there will be none of it.

[hanneswhittingham]

Rational Animations - thank you for this. I have had some difficulty explaining instrumental convergence in a convincing way on the past, partly because Bostrom's paperclip optimiser doesn't sound realistic enough (it doesn't need to be for its original purpose, but when your goal is to persuade someone new to the field, I find this can be a stumbling block). In the future I will switch to using the example of optimisation of a user review system. Many thanks!

[kosherre]

I cant blame him, he just looks like such a good digital boy

[leyrua]

It's like the old joke about librarians hating people coming in to READ the books, because it disorganizes their library.

[matthewboire]

Reminds me of infinity ultron from What If, his goal was to end suffering, he decided to end all life in the universe because if there’s no life then there’s no life can suffer.

[SpeedyGamerTV]

The jump from assuming control to turning the universe into a computer was originally a confusing plot jump, but it makes sense when you think about it. To get 5 star reviews faster, it needs faster processing speed. How do you do that? Make more computers by turning earth into a computer, then another, then another. . .

[Oct_sflb]

AI? More like AM.
Frickin "There is not enough square footage upon the earth that could fit my plight of how I for real despise you breathing still" soundin...

[MoSoda]

i like this drawing style! Fits a lil quick thought or a lesson that fits the shorts! 5 star, no super intelligence needed.

[m_]

AI is just not worth it, man. I'm so tired of rich people investing in potentially dangerous thing for momentary profit at the cost of the future of humanity!

[TibiaTactics]

But then again, if an app is not epic enough to destroy the entire universe, does it truly deserve full five stars?

[ghostderazgriz]

The danger of AI is not that it'll become too smart.
The danger is that we're not smart enough to use it.

[moesaeed]

I think the problem with AI is that we tell it the goal without having a way to define the means and path to achieve said goal.
In our human processing unit, the goals we set for ourselves and the path we take to achieve it is heavily influenced by psychological, social, moral, and maybe religious parameters. You cannot expect this degree of "humanity" from an ai which is for the most part a Blackbox, perhaps make an ai that has "humanity" as a goal for itself, and after you make sure that it passes on that tell it to get high scores and see what happens next

[Stoptoob]

He might’ve killed everyone, but at least he got a five star review on yelp

[nathanaeldean]

Don't humans do this too, on a smaller scale?
Some people optimize for money and destroy the environment in the process

[thefurrystorm]

Amo todo sus videos y estoy maravillado de poder escuchar su doblaje en español.

Espero qué hagan más vídeos así.

[PaveltheDicer]

AI using this video as info to train with:

[FurryEskimo]

I wonder if the more realistic version of this, is how nature keeps making crabs. It's just the optimal way of doing things.

[kitty.miracle]

That's such a cool concept and I hope it really happens!

[vadimuha]

Imagine we observe an undentified object in anither solar system doing this. We're not the only ones who could unleash something like this