AI safety is being dominated by concerns about AI takeover and human extinction. They want you to believe they are sounding the alarm on "the danger". But there is no singular danger. There are dangers, and every plan has tradeoffs. No one has a monopoly on safety.

Oh yay the EA safety people are community noting this because of a joke I made in the video because I said “upon the vesting of my shares” and apparently this means I said his Anthropic shares vested? My point is just “dude has already benefited tremendously from his time at the labs that are apparently going to kill everyone”. My god I hate this platform sometimes.

this account is so incredibly underrated.

My haters would like you to know they think I am overrated

We need more voices of reason to combat the hysteria

"The lack of diversity of views terrifies me." 100%

It is so scary how people think there’s this one risk and that’s what we need to solve for

I'm so pissed off - what other industry could be like "hey we need to slow down because danger you're welcome" And the world goes thank you my wise kings. unfathomably gross

🤣 Love the delivery! Like, that’s what you believe and you are like, yeah I support that guy for standing up. I’ll see you at work on Monday.

Hahaha ya mad respect

The EA hysteria reminds me of this scene: “It would be 9/11 times a hundred!”

Hahaha, forgot about that too

You have been incredible throughout this. Thank you for saying what I'm sure many industry veterans are thinking. Please don't give up, there aren't many who have your speaking ability who haven't take one of the easier, more extreme sides.

Really appreciate it! I’m trying but damn it’s demoralizing at times haha

I watched this video right *after* making a very similar speech to friends. You said it better than I ever could. I am demoralized too.

The issue is also overestimation of the labs’ role. The conflict between the labs & the WH reminds me of the film when Truman tells Oppenheimer “You think the people in Nagasaki & Hiroshima give a shit who built the bomb? They cared who dropped it. I dropped the bomb, not you.” No one in Yemen or KSA cares who built Claude. They only care about the missiles being built

Yea but the labs think they’re also going to be the one to drop it

Dwarkesh referred metr as well it’s crazy they are saying it out loud

They want metr so badly

Anthropic and OpenAI employees edging the end of the human race is such a fucked-up kink that I might actually shame it.

Calling out @DarioAmodei and @sama bullshit love it! 🤣

@DarioAmodei @sama Oh no now Dario might see this and then he will send his ai to kill me

@DarioAmodei @sama Oh no, I really shouldn't have tagged them I hope they don't see this. Praise Dario and Sama! 🙌

@DarioAmodei @sama I take it back, wut

@DarioAmodei @sama lol the insane part about this is the implication that if the us said we are nationalizing he might be like “no”

@DarioAmodei @sama the most entertaining outcome is the most likely 🤪

They are playing doomers like a fiddle. Notice they have been saying this since ChatGPT 2, and everyone knows that wasn’t world ending The truth is that they don’t have the compute to release the larger models. They are burning money, and need to wait for compute to come online

Thank you 🙏🏼

Hah, trying my best!

Effective Altruists trying to figure out how many holocausts AI will create:

AI deaths will be ~271,000 at best

What is 271k?

That was a holocaust joke, not really important. The paperclips and shrimp welfare are what we must focus on instead

excellent.

Great points as always but ngl the lighting is very pleasing

Hahaha thanks. I’m still working on getting it right, so probably will keep changing it up

People are doing...

I will commit crime and i want my friend to be the judge for the case.

If the gov would just indict someone (anyone or a company) for a felony (or even a class A misdemeanor) related to the CFAA then all this yolo AI safety shit would add AI security real fast.

Christ Almighty dude, the bird hands! 🤢🤮

Those are just my normal hands!

It isn't a chance they will kill 10% it 10% they kill all 100% of humans.

I know, but in expectation that’s the same thing. 10% chance of 8bn deaths is an expected 800m

Expectation is for average outcome. Not meaningful for a singular binary event. Either everyone survives or all do not. That is the claim being made. Not some AI might kill some humans.

Couldn’t have said it better myself - the biggest danger of super intelligence is concentrated ownership by the few who can afford it and leaving the rest of us behind. Just like there are no guarantees AI destroys humanity, there are also no guarantees we get to share in its prosperity.

Yea exactly that. And people are pushing bad on the “we don’t get to share in the gains” path to avoid the “everyone dies” path when the latter is far more speculative.

If we all going to die one way or another let’s all agree on a way. I vote for use AI and John Hammond to create Jurassic Park and bring back dinosaurs 😂 I’ve always wanted a pet raptor anyways. Fuck that guy in his kitchen