Elizabeth Van Nostrand 'I'd rather use well defined, short, jargon than the social sciences' habit of using common words with a new definition.'

As would I! But the jargon could very often use building blocks that laypeople are able to understand. There's no ambiguity or namespace collision from…

I wonder what the annual death toll is from Anglophone scientists preferring words like 'murine' over 'mouse-ish' in order to sound fancy, thereby blocking lots of laypeople from understanding or remembering things like medical terminology?

Researcher: jobs.ashbyhq.com/miri/c5a85cd2-…

Writer: jobs.ashbyhq.com/miri/44b7a3a1-…

The roles are located in Berkeley, and we are ideally looking to hire people who can start ASAP.

Please share this with your networks or any people you think might be a good fit!

MIRI is hiring for roles on our new Technical Governance team! Links in thread!

We're looking for 2–4 Researchers, and 1 Writer, to help us research and design technical aspects of regulations and policies for addressing the risks associated with smarter-than-human AI.

michael vassar 🐦🔬🔭🌴🕺🏋️‍♂️🧘‍♂️🤸‍♂️🧍‍♂️👨‍👩‍👦 Zvi Mowshowitz Sarah Constantin I meant to suggest, like, read a science paper about a known issue, think about it for a few minutes and devise a plausible counterspell, then go back to doing things and see if the counterspell helps.

Hey there's only two more days to apply to MATS. If you're a technically talented person and have some interest in trying out alignment research, I strongly recommend applying, and I don't even work there anymore. Applications close the 7th. Link in replies.

a common rat/ssc/tpot mistake is reading charitably by mere habit, not as a thoughtful decision

if youre trying to have a useful conversation w someone, be charitable with their words

if youre trying to understand what they actually think, charity isnt appropriate

Published in Science, coauthors include a Turing Award winner (for co-inventing deep learning) and a co-author of the world's leading AI textbook:
science.org/doi/10.1126/sc…

'How do you think AI systems can do so much? You must not understand them well enough to see that there's no magic in them'

Yeah, magic isn't actually required. You have to understand the rest of the world well enough to see that there's actually no magic anywhere else either

They think we're making the mistake of not applying reductionism to AI, but actually they're making the mistake of not also applying reductionism to everything else

Everything is 'just' the stuff it's made of

'AI is just linear algebra' feels like a midwit thing, where someone's figured out enough about the world to be excited by reductionism, but not enough to realise that it's universal.

This is with my excellent co-authors Noam Kolt, Yoshua Bengio, Gillian Hadfield, and Stuart Russell.

See the paper for more discussion on the particular dangers of long-term planning agents.

What should governments do about this?
science.org/doi/10.1126/sc…

This is connected to Matthew Yglesias's argument here; you can see the smartphone panic in a continuum with prior anxieties about the effects of screens and also see smartphones as the moment when those effects started *really* vindicating the basic anxiety.
slowboring.com/p/is-ever-bett…

Recent research justifies a concern that AI could escape our control and cause human extinction. Very advanced long-term planning agents, if they're ever made, are a particularly concerning kind of future AI. Our paper on what governments should do just came out in Science.🧵

Claude has also demonstrated it can guess when it is being tested. (That's not to say that in this case, it used that information to decide to pause any misbehavior). twitter.com/alexalbert__/s…

Well, we shouldn't allow such AI systems to be made! They haven't been made yet.

A key problem with sufficiently capable long-term planning agents is that safety tests are likely to be either unsafe or uninformative. Suppose we want to answer the question:

Hominids were trained by evolution to chip handaxes. Humans went to the moon.

You don't only get what you trained for, when the thing you end up training is intelligence.