Benjamin Henke

Google DeepMind wants to know if chatbots are just virtue signaling We need to better understand how LLMs address moral questions if we're to trust them with more important tasks.

MIT Technology Review coverage: www.technologyreview.com/2026/02/18/1...

A roadmap for evaluating moral competence in large language models - Nature This Perspective offers a roadmap for tackling the challenges of the facsimile problem, moral multidimensionality and moral pluralism in large language models.

New in @nature.com: We must move beyond mimicry to assess AI for genuine moral competence. We propose a roadmap for the ‘facsimile problem’ that accounts for moral multidimensionality and pluralism—a path toward more responsible AI.

www.nature.com/articles/s41...

Important work from @birchlse.bsky.social

1. It was the nightingale, and not the lark,
that pierced the fearful hollow of thine ear.

2. It was the nightingale, and not the lark,
that pierced the fearful hollow of thine ear.

Can you REALLY tell the difference between AI-generated and human-written text? One of these texts was written by a human and another by a well-prompted chatbot. Which is which?:

‘Gal’ comes to mind. I mostly wish it didn’t.

Next up in our ongoing AI Affect series:

👤 Tim Salomons (Queens Canada)
📢 "How do we judge others' pain?"
🗓️ TOMORROW April 15, 3-4:30 PM
📍 Join us in person or online! DM for details.

Responsible AI

Announcement of Institute of Philosophy Conference on Responsible AI at the University of London on 19-20 May. Free and open to all

philosophy.sas.ac.uk/news-events/...

🙄

But her emails

Conversing in the Dark: Off-Off Record Speech Acts and the Cooperative Creation of Uncertainty | Sam Berstler (MIT)

Join us Friday for a bonus PPE tak!

Sam Berstler (MIT) | "Conversing in the Dark: Off-Off Record Speech Acts and the Cooperative Creation of Uncertainty"

📆: Fri, March 28, 4:30-6 PM
📍: Senate House, Rm 349

Hope to see you there!
philosophy.sas.ac.uk/news-events/...

Next up in our ongoing AI Affect series:

👤 Rob Long (Eleos AI)
📢 "Taking AI Welfare Seriously"
🗓️ Tuesday, March 25, 3-4:30 PM
📍 Join us in person or online! DM for details.

Ride out with me!

Next up in our ongoing AI Affect series:

👤 Tom Everitt (Google Deepmind) @tom4everitt.bsky.social
📢 "Agency as backwards causality"
🗓️ TODAY Tuesday, March 4, 3-4:30 PM
📍 Join us in person or online! DM for details.

The deadline for applying to be an AI fellow at the LAIHP is this Sunday.

Two cartoon trolleys are talking while another trolley, called Kyle, is about to run people over behind them. One says, “I'm concerned about Kyle.” The title reads: “Problem Trolley.”

A cartoon by Amy Kurzweil. #NewYorkerCartoons

I know you’re wrong, but I’m just having no trouble articulating why.

Developer creates endless Wikipedia feed to fight algorithm addiction WikiTok cures boredom in spare moments with wholesome swipe-up Wikipedia article discovery.

It's a neat way to stumble upon interesting information randomly, learn new things, and spend spare moments of boredom without reaching for an algorithmically addictive social media app.

University College Dublin is hiring five year Ad Astra Fellows in the Philosophy of Technology/AI. Deadline: February 21st. Find out more at the link below.

Next up in our ongoing AI Affect series:

👤 Jeff Sebo (NYU) @jeffsebo.bsky.social
📢 "The Moral Circle"
🗓️ Tuesday, February 4, 3-4:30 PM
📍 Join us in person or online! DM for details.

FYI, I just got followed by this account: bsky.app/profile/laur...

How has DeepSeek improved the Transformer architecture? This Gradient Updates issue goes over the major changes that went into DeepSeek’s most recent model.

Very good (technical) explainer answering "How has DeepSeek improved the Transformer architecture?". Aimed at readers already familiar with Transformers.

epoch.ai/gradient-upd...

The LAIHP is pleased to invite applications for Visiting Fellowships at the Institute of Philosophy, School of Advanced Study, University of London.

The fellowship period will run from May 19th-June 27th, 2025

For more information, see below.

Today, we are publishing the first-ever International AI Safety Report, backed by 30 countries and the OECD, UN, and EU.

It summarises the state of the science on AI capabilities and risks, and how to mitigate those risks. 🧵

Full Report: assets.publishing.service.gov.uk/media/679a0c...

1/21

This isn't an answer to your question, but a strong relationship between desireableness and credence would make sense if we adopt an RL view of desireableness. A surprising positive result is more reinforcing than an unsurprising one.

Reminder @alex-taylor.bsky.social is speaking *this* Thursday discussing his BRAID project on red teaming & outsourcing labour in the Global South.

Make sure you don't miss out - get your hybrid ticket now 👉 rb.gy/64oljm

@technomoralfutures.bsky.social @edcdcs.bsky.social @uoe-gail.bsky.social

Can't wait for this talk tomorrow?

Too bad, that's when it is. We're excited too.

This is an evidence-free space. Please leave.

Or, more correctly, a-lu-mi-num. spelled accordingly

Wait, brits SPELL it aluminium?