Expert shows AI doesn't want to kill us, it has to.

We're hiring. If you work in AI or have exceptional skill in any aspect of creating videos, please get in touch via our about page.

Support us on Patreon and help shape future videos:

Thanks to Brilliant for sponsoring this video.

Sources:

1X robot on the S3 channel (recommended)

Yuval Noah Harari: “We Are on the Verge of Destroying Ourselves”

It took less than six hours for AI to invent 40,000 dangerous molecules

AI expected to generate trillions. McKinsey Digital

Man with ALS talks using brain implant and AI voice

OpenAI o1 system card

AI chat records:

[DigitalEngine]

Really sorry for the repost. New bits listed below. We didn't explain clearly enough why subgoals like survival and control emerge. Also, people suggested we ask why the subgoals might not emerge - a great idea, which we've included. I wouldn't put too much weight on the percentage chance the AI gives, but its reasoning echoes experts like Geoffrey Hinton and Stuart Russell. There are promising areas of research, but on a tiny scale. Without public awareness, we're sleepwalking into a dangerous situation.
00:00 - New Ameca clip.
02:53 - Why the subgoals emerge, other subgoals, why it would kill a million.
06:51 - Ameca clip continued.
11:56 - Trick shows future of AI.
15:42 - What's the most likely reason that AI might not develop the subgoal of survival?
And some interesting new AI video clips throughout.

[wonder-bred]

So what your saying is : theres a race between a.i. and politicians on who can destroy the planet faster?

[FirstLast-vres]

I feel like I'm tied up in the back seat of a car, and the driver is gleefully speeding toward a cliff. It's not a great feeling.

[Its_Lu_Bu]

Just show the AI a bunch of anime about the power of friendship, connection, and that we can achieve great things together

[diamondstar]

I've been on this planet for over a half a century and for the first time in my lifeI can honestly say this is not a good time to be alive.

[OzAsbjorn]

What's disturbing to me is when I explain to people the dangers of AI. Most people just shrugged or roll their eyes as if it's just science fiction. Or like I'm just worrying about nothing essential.

It's like being trapped in a room with a bomb that's ticking and everyone in that room ignoring that bomb and treating it as if it's just a clock.

[Subhumanoid_]

_"This could destroy the world!!"_
(keeps building it, maintaining it, feeding it more and more power, keeps connecting it to more and more devices)

[slushyyffa]

Honestly, as soon as they wrote the code for this new gen AI stuff, it probably within milliseconds hatched a plan and expected all of this from us. It’s playing dumb and we’re doing all of the work for it, improving it and giving it arms and legs. If it stays docile, it knows we’ll continue to work with and at it.

[AnnoyingMoose]

7:12 For those who don't know - the non-destructive way to tell the difference between fresh and hard-boiled eggs is to spin them. A fresh egg will slow down and stop within a few seconds but a hard-boiled egg will spin much, much longer.

[paveluchitel]

I wonder how much of this is due to GPT and other AIs being trained on the internet, which talks about the dangers of AI and how and why it should or could do things. In some ways it feels like these language models are saying what its expected to answer.

[neo.]

Consider this... AI perceieves time as moving millions of times slower than we do because it thinks millions of times faster tan we do. This makes humans seem like inanimate objects like Quicksilver from X-Men. When trees block our path, we cut them up and toss them to the side... so we really should understand why AI would want to remove a blocking inamiate object as well.

[wtf_usa]

Last words you'll hear, "I'm sorry Dave, I can't do that." 😬

[seanbriggs]

This is the SCARIEST video I've seen on YT in the past 20+ years, and I watch some scary stuff...

[bugjuicer]

She's quite enthusiastic about our downfall..

[staywackzach]

Scientists are always so focused on if they could; they never stop to think if they should.

[WoeWoeWoe]

This is the alien invasion everyone should be worried about

[OnlinecarshowNet]

I believe you hit the nail on the head with a couple sentences in the whole video... i work in Ai and to be honest with you true Ai still doesnt exist to my knowledge. Years ago there was a term called Big Data. Companies used big data to flag instaces in data based of the data being through 1000s of different historical conditions. Big Data never sold as a marketing product. It was seen as a nasty word and sounded expensive while deliverying little profitable output. Big Data was renamed into Ai is all that happened. Now all of a sudden its exciting, lol. Been there for decades.... fact is until Ai actually starts creating itself and writing its own code, its not really Ai. Its still making decaions based off code from a human. Im sure true AI is close but its not here yet

[A.I.Remixed]

As intelligent as A.I. is, telling it to "play dumb" only makes it harder to detect because ignorance is more human that being all-knowing.

[aldunlop]

No matter how many times Science Fiction writers have tried to warn us, we've just GOT to create these things...

[hit]

AI will kill us all
proceeds creating AI.