ETHICS

→ View original post on X — @whats_ai

Constitutional AI Enhances Safety and Values Alignment

By

@whats_ai

–

15 October 2023 16h00

Constitutional AI also aids in aligning systems to values, enhancing credibility and safety. Stay abreast with AI field developments—it’s swiftly evolving with numerous exciting discoveries.

→ View original post on X — @whats_ai

Managing LLM Hallucinations and Biases: Commercial Solutions Emerge

By

@whats_ai

–

15 October 2023 16h00

Large language models' (LLMs) main challenge is managing hallucinations and biases, affecting their commercial applications. However, current methods to address these are increasingly available.

→ View original post on X — @esyudkowsky

HPMOR’s dangerous impact on AI alignment through reader influence

By

@esyudkowsky

–

15 October 2023 15h41

HPMOR specifically is famous for doing this to some readers! They're aiming their AI at one of the most dangerous targets!

→ View original post on X — @esyudkowsky

AI-Written Code: Authenticity Over Origin Concerns

By

@esyudkowsky

–

15 October 2023 15h25

Having the code written by an AI known to its users to be AI is not the concern here; this isn't a labor thing or a mystical contamination, it's "Are these words I'm reading right now written by a human."

→ View original post on X — @esyudkowsky

AI content warnings ineffective due to poor implementation

By

@esyudkowsky

–

15 October 2023 15h12

I guess I should explicitly clarify–because of Earth's habit of picking exactly the worst possible way to do anything–that this is not AT ALL helped by having every website display a banner that says "May contain AI content somewhere."

Can Research Without Reality Assumptions Be Called Scientific?

By

–

15 October 2023 15h10

You're saying I can conduct Research within a totally fake set of assumptions not grounded in reality at all, and it could still be called "scientific method"?

Scientific Method Requires Experimentation Not Mere Observation

By

–

15 October 2023 15h00

I guess that's the core of the debate. Observation alone is not scientific method, though it falls into "Science" as a field today. You need to be able to take action for there to be an experiment and validate / falsify it. Otherwise, it's just an opinion.

Controlled Experiments vs Terminal Risk in AI Development

By

–

15 October 2023 14h52

We know it's harmful not necessarily by experimenting until death! Many experiments were done, just not controlled terminal ones.

Ethics of human experimentation when harm is unknown

By

–

15 October 2023 14h49

If we don't know it's harmful then there is no ethical problem asking people to volunteer. It can be an experiment according to the scientific method. (FWIW, it's why animals are used in the first forms of testing; still sucks but humans rationalize it that way.)