Aaron Maiwald

Make the future awesome

[James Black @jrm_black]

New paper. In biosecurity, we're worried about being able to efficiently generate novel genomes. We consider two axes, evolutionary novelty and design efficiency, for genome design, applying this framework to Evo-designed phages. Our analysis suggests we’re not there yet ⬇️🧵

[Pivotal Research @pivotal_org]

Applications open for the Frontier Biodefense Fellowship!
Work on high-impact biodefense projects in London for 9 weeks.
Mentors from SecureBio, Coefficient Giving, Blueprint Biosecurity, American Enterprise Institute and more.
Stipend, travel, accommodation, computer, and a dedicated desk at LISA are all covered.

Who is it for? Anyone committed to working on biodefense!

📍 London • 📅 Aug 3 – Oct 2 • ⏰ Apply by Jun 2

[Jack Morris @jxmnop]

i hate it when the new claude code starts using tools and googling stuff. like that's my job. i google stuff and then paste it to show you what i saw. don't leave me out of this

[Bulletin of the Atomic Scientists @BulletinAtomic]

Tech needed for tracing engineered biothreats back to their sources are advancing rapidly.

@OllyMCrook, @AnemoneFranz, and @MaiwaldAaron share recommendations for creating attribution systems that can improve detection, accountability, and deterrence: https://t.co/4EkDEljx6D

[David Sinclair @davidasinclair]

What a powerful video

[Kevin K. Yang 楊凱筌 @KevinKaichuang]

Structure - > sequence models have different phylogenetic (and biochemical) preferences than protein language models.

@OllyMCrook

[Jan Brauner @JanMBrauner]

Out in Science today:
In our paper, we describe extreme AI risks and concrete actions to manage them, including tech R&D and governance.
“For AI to be a boon, we must reorient; pushing AI capabilities alone is not enough.”

[Aaron Maiwald @MaiwaldAaron]

@jmschreiber91 @drklly I'm using integrated gradients on a BERT model and the convergence deltas are basically all > 0.01 and mostly > 1

[Aaron Maiwald @MaiwaldAaron]

@jmschreiber91 @drklly Great thread, thanks!

How close to zero do you think the convergence deltas should be (roughly)?

[Tyler is finishing a book, slow to reply @TylerAlterman]

For the benefit of all sentient beings

[Aaron Maiwald @MaiwaldAaron]

Sincerely wishing people well in your thoughts when you walk past them, is a beautiful small way to become a kinder person

[Aaron Maiwald @MaiwaldAaron]

@teebarnettOG just listening to your podcast with Lulie and its really, really good! I used to host a podcast as well and I'm really impressed by your ability to guide the conversation!

[Aaron Maiwald @MaiwaldAaron]

@LuiseWoehlke So beautiful!

[Aaron Maiwald @MaiwaldAaron]

@MatthewJBar More generally, once you jailbreak GPT-4 (which is definitely possible at the moment), it become a very competent, general-purpose crime assiste t

[Aaron Maiwald @MaiwaldAaron]

@MatthewJBar I don't think GPT-4 is a harmless tool at all. Have you read this paper?

https://t.co/7Dk7ZaztYN

[Aaron Maiwald @MaiwaldAaron]

@robertwiblin @tobyordoxford i'd also be really interested to hear his views on which approaches in technical alignment and compute governance are most promising

[Aaron Maiwald @MaiwaldAaron]

@robertwiblin @tobyordoxford what are the main ways in which his x-risk estimates have changed since writing 'the precipice'? in particular, how and why has his x-risk estimate from AI changed?

[Aaron Maiwald @MaiwaldAaron]

@AndrewYNg Thank you, I'm really grateful that you are thinking about these questions!

[Lewis Bollard @Lewis_Bollard]

Progress! 1,157 companies globally have now fulfilled their pledges to stop using eggs from caged hens.

Over 200 million animals are today out of cages thanks to this progress.

I'm very grateful to the advocates whose hard work has achieved this.

https://t.co/K6qLUv6Feo

[Aaron Maiwald @MaiwaldAaron]

@anthrupad @ohabryka @ESYudkowsky Would be interested too