Sebastian Benthall @sbenthall

Research Scientist at @ILI_NYU and @ICSIatBerkeley. A dad. Tweets are reflex, noise, and error. 🦋https://t.co/t1n5c7A8Tp digifesto.com An NYC suburb Joined April 2008

Tweets

16K
Followers

1K
Following

657
Likes

14K

Sebastian Benthall @sbenthall

a year ago

I'm building something new. digifesto.com/2025/05/28/im-…

0 1 2 113 2

View Details

Sebastian Benthall @sbenthall

2 years ago

What about the loyalty of AI agents in government function? digifesto.com/2025/02/06/wha…

0 0 0 102 0

View Details

Sebastian Benthall @sbenthall

2 years ago

Complex systems and ethics according to contextual integrity digifesto.com/2024/12/08/com…

0 0 0 84 0

View Details

@AndrewCritchPhD There's a fair case to be made for all anthropomorphic self-references of contemporary chatbots to be deceptive, but we might need a new kind of grammar for the speech of products and services.

0 0 1 9 0

View Details

Sebastian Benthall @sbenthall

2 years ago

@AndrewCritchPhD 'Person' is a much more capacious term than 'human', as we already have artificial persons (i.e. corporations). At the policy level, the question is going to be about the relationship between corporate persons and robots pretending to be people but deployed by corporations.

1 0 1 20 0

View Details

Sebastian Benthall @sbenthall

2 years ago

a brief reflection on where I'm at as a scientist digifesto.com/2024/11/26/how…

0 0 1 93 0

View Details

Sebastian Benthall @sbenthall

2 years ago

Anyone had luck with Bluesky yet?

0 0 0 60 0

View Details

Sebastian Benthall @sbenthall

2 years ago

Ok, but what is the anti-incumbency wave about? Is there an explanation of that yet? Is it just 'inflation'?

1 0 1 122 0

View Details

Sebastian Benthall @sbenthall

2 years ago

@zephoria Will you be in Ithaca or NYC?

0 0 0 45 0

View Details

Sebastian Benthall @sbenthall

2 years ago

@Miles_Brundage Seems like the original sin of the AGI definition is that it's based on a comparison with a human as a unit of cognition. What would you say are better categories/concepts to use? I'd welcome recommended reading on this.

1 1 3 338 0

View Details

Sebastian Benthall @sbenthall

2 years ago

@AJA_Cortes parenting

0 0 0 4 0

View Details

Nora Belrose @norabelrose

2 years ago

This is a great paper. It points out: 1. Humans do not even approximately behave according to rational choice theory 2. There is no reason to think advanced AI will "inevitably" maximize some utility function 3a. Human preferences are derivative / constructed, so aligning AI by matching its behavior to our stated preferences is wrongheaded 3b. We can align AIs directly to some normative ideal of a "good assistant / programmer / driver / etc." instead 4. Aggregating preferences across people is fraught with philosophical and mathematical difficulties. We should not aim to align AI to the "collective will of humanity."

xuan (ɕɥɛn / sh-yen) @xuanalogue

2 years ago

Should AI be aligned with human preferences, rewards, or utility functions? Excited to finally share a preprint that @MicahCarroll @FranklinMatija @hal_ashton & I have worked on for almost 2 years, arguing that AI alignment has to move beyond the preference-reward-utility nexus!

34 169 985 284K 862

39 137 978 159K 844

View Details

Sebastian Benthall @sbenthall

2 years ago

Direct, concise answers to coding questions in Google AI Overview is a real improvement, saving a few clicks every half hour or so.

0 0 0 98 0

View Details

Sebastian Benthall @sbenthall

2 years ago

I miss writing. digifesto.com/2024/10/03/i-m…

0 0 1 54 0

View Details

Sebastian Benthall @sbenthall

2 years ago

I had a great time at the 6th Annual Symposium on Applications of Contextual Integrity! It's a great community that's always growing intellectually, with a sincere and friendly vibe. Thanks to everyone who organized it.