AI Ethics

Context

Can an AI be ethical?

Is that like asking whether a machine can be ethical?
Clearly a machine can create a legal obligation.
So, let's agree that it is unethical to deploy an AI that is not aligned with human ethics.

• What Is the Alignment Problem?

At its core, the alignment problem asks:
How do we ensure that AI systems pursue goals that reflect human values, intentions, and safety—especially as they become more autonomous?
It’s not just about avoiding catastrophic outcomes. It’s about embedding agency, consent, and accountability into systems that increasingly act on our behalf.

• Ethics: The Heart of the Problem

Alignment is fundamentally ethical because it involves:
Value selection: Whose values are encoded? Whose are excluded?
Moral pluralism: Can one system reflect diverse, even conflicting, human norms?
Consent and autonomy: Are users aware of what the system is doing? Can they opt out?
Responsibility: Who is accountable when things go wrong?

Evaluation

https://www.civics.com/evals

Luis Felipe Murillo and Sara E. Berger, Meta-Eval:Toward a metapragmatic approach to large language model evaluation https://nextcloud.unixjazz.org/s/NgmWNi2Jm3jojR5

2025-09-05
Mark Novak - I still think that least privilege is the way. So one agent finds the movie, another spends the money: up to $4.40, only using Prime Visa, on Amazon only. The first calls the API on the second.
Tom Jones  - this is a different architecture that i was imagining.  I was thinking that i have one session connection to one agent and then that agent reaches out to others. You seem to be imagining a multirooted mesh - if i am not mistaken?
Mark Novak - I’m not sure how it’s going to shake out, but I am imagining that agents taking on various roles and permissions dynamically would be very difficult to control and prone to giving end users all sorts of headaches.