Re: AI ethics [was: The Truman Show]

Michael Nielsen (
Mon, 22 Jun 1998 18:09:12 -0600 (MDT)

On Mon, 22 Jun 1998, Alexander 'Sasha' Chislenko wrote:

> At 10:31 06/21/98 , Michael Nielsen wrote:
> >
> >Is it ethical to contain an AI in a limited world? This is an especially
> >interesting question if one takes the point of view that the most likely
> >path to Artificial Intelligence is an approach based on evolutionary
> >programming.
> >
> >Is it ethical to broadcast details of an AI's "life" to other
> >researchers or interested parties?
> >
> >Is it ethical to profit from the actions of an AI?
> I would suggest two interpretations to the above questions.
> Human ethics evolved to regulate people's relations with other persons
> and their possessions in the physical world. Some of the ethical
> principles must be generic and good for regulation relations between
> different entities (the degree of identity granulation in an AI world will
> be much different from human, so I don't really expect "personalities"
> there). Other principles may be just carried to the new world, because
> we "feel like it". I think this is a wrong approach. Should we cover
> software agents with a decency crypto-veil in the moments of replication,
> or tax successful computer programs to provide a decent level of
> operation to their underprivileged inefficient brethren? This can get
> quite ridiculous.

If an AI told you that it wanted privacy to perform a particular
operation, would you respect that desire, even if it bordered on the
apparently ridiculous?

> We should better think of how to efficiently set up new systems, and
> what kinds of protocols may provide us - and AIs - with greatest mutual
> benefits.

A very interesting problem to consider.

> If there are any ethical controversies with AIs, they will probably happen
> only in the short period of time (10 to 20 years?) when AIs have near-human
> intelligence.

I don't doubt that AI is going to occur, assuming the human race doesn't
undergo a radical change of orientation in the next few decades, but I'm not
that optimistic about the time frame. The only argument I've seen for
that is Moravec's, which seems to consist largely of hand-waving.

Michael Nielsen