[ExI] Ethical behaviours

[Ben Zaiboc]

BillK <pharos at gmail.com> wrote:

> If you design an AGI using cognitive functions only ...


I'm pretty sure that's an oxymoron.

Although, it depends on what you mean by 'cognitive functions'.

I don't like the term "AGI" myself, but I understand it to indicate an AI with a broad problem-solving capability, similar in type to (but probably greater in scope than) human abilities.  This implies a lot more than just calculating ability, but includes emotions, imagination, theory of mind, and so-on.  In other words, an artificial person, capable of acting as a moral agent.  There may be human examples of smart, unscrupulous people, but I suspect that a sufficiently advanced AI would have superior morals as well as superior intellect.  

The fact that there are morally confused intelligent humans isn't really that relevant.  The difference between the dullest and the smartest humans is tiny compared to the sort of intelligence we are expecting machines to reach, so I don't think it makes sense to hold up clever bankers as evidence that intelligence and morality are completely separate.


Ben Zaiboc