Jonathan Mannhart 🔎🔸

Third on this year's Black List (the most liked scripts that have not been made into movies yet), this:

What??

(Yes, this is real: variety.com/2024/film/ne...)

YouTube video by Kungl. Ingenjörsvetenskapsakademien IVA Seminar with Professor Geoffrey Hinton, at the Royal Swedish Academy of Engineering Sciences (IVA)

Seminar about AI with professor Hinton at the Royal Swedish Academy of Engineering Science last week. We discussed AI safety, intelligence, embodiment, unpredictability, and many other things. youtu.be/lexF-CrhOrE?...

It’s an alivist bias that stems from an extraordinarily privileged position.

The sheer arrogance of demanding security in extending the temporal reach of one's own entropy-producing activities. When literally every single ancestor followed the natural convention.

Real x-risk means that a superintelligent AI wants to end biological life in every instance of the multiverse, what we're dealing with is just a scary sounding local limit of survivability.

LeCun, 2026:

“It's highly plausible that accidental AGI misalignment due to goal misspecification can lead to asymptotically everybody dying. But that doesn‘t mean that it's the optimal way to achieve human extinction!“

Thirteen years ago today, I donated a kidney to a stranger, starting a chain w/ 6 recipients. It felt like a great deal: with modest risks, I could give someone many more years of healthy life. Looking back, I'm glad I did it, but I understated both the risks and the benefits. 🧵

Have a look at the authors! Whoa!

Extraordinarily excited for this book by @dwarkesh.bsky.social and @gleech.org

The Scaling Era: An Oral History of Large Language Models The Scaling Era: An Oral History of Large Language Models [Patel, Dwarkesh, Leech, Gavin] on Amazon.com. *FREE* shipping on qualifying offers. The Scaling Era: An Oral History of Large Language Models

coming soon...

www.amazon.com/Oral-History...

So if GiveWell recommends the Against Malaria Foundation, and Giving What We Can (@givingwhatwecan.bsky.social) recommends people give 10% of their income to effective charities, that makes a lot of sense to me.

And it's likely (even) more effective than taxing billionaires.

What Effective Altruism (in my opinion) does well, is that those people really think about what the *most* neglected areas and most neglected people are.

The extremely poor (Malaria), animals (ending factory farming), etc. are things that most states don‘t do that much for.

This is not an argument against higher taxes in the US. It's an argument for why I don't think it's necessarily true the state is the most effective way to help the ones most in need. Especially an extremely rich western state.

The US state (or the German state, where I live), doesn't exactly do what I would want my taxes to do. I think the US prioritises US citizens to a degree it shouldn't, esp. under a Trump administration.

And deprioritises the global poor.

I prefer the Against Malaria Foundation instead!

How does it do that?

I donate 10% of my income to the Against Malaria Foundation (which, according to GiveWell, gives me probably the best bang for my buck).

I do that because of EA people & ideas. And because Malaria fucking sucks and we should end it.

What's your problem with me doing this?

Electric bikes/cargo bikes are much safer than motorbikes/cars:

• Speeds < 30-40 km/h
• Low center of gravity
• Better situational awareness—360 visibility and low noise
• Pedal-assist prevents rapid acceleration accidents
• Much lighter mass (20-40 kg)
• Plus exercise, low-emissions and fun!

Civility? Not you too, Paul, no!! I don't want to have to block you if you keep going down that road

textbook Vlad, such a him thing to say. Totally giving vlad vibes, 100% unfiltered just like I knew him

Underrated. Rice and tofu can already both be delicious just as they are, many don't know this.

It took the Japanese hundreds of years to figure out that just putting a slice of tuna on some rice is a good dish, and then the whole world went crazy for it 20y ago!

Somebody needs to write an article criticising that reply while misunderstanding it, otherwise it doesn‘t really count

This is just surprisingly fun to play around with

It appears that ChatGPT is currently experiencing issues with the name “David

I'm so glad Biden pardoned the writers for adding that fantastical element

I can't believe they made a whole movie inspired by alien intelligences built on complex self-attending architectures—and then found out they scale so well they now make a new iteration every couple of years.

Wow!!!

Q* (i.e., Strawberry) was inspired by the dangerous technology and international espionage and intrigue in the movie “Oppenheimer“! There is no way!!

Such a good movie too. Imagine if a dangerous situation like that really happened!

X community note made of all Bluesky links.

:)

Also see: bsky.app/profile/vict...

(I.e. if that is the source for the discussion in the podcast then it was just an error.)

This has already been discussed, if you click through to the original post: bsky.app/profile/tilm...

☝🏼 thanks for linking the discussion!

Just clarifying that the movie 'Arrival' (based on Ted Chiang's 'Story of Your Life') wasn't a direct inspiration for self-attention.

It's a tempting meme though since the analogy is not bad & has that 'sci-fi predicts/creates reality' theme

Sigh.