Re: Rights for AIs are not a problem

From: Brian Atkins (
Date: Mon Jan 16 2006 - 20:22:15 MST

Keith Henson wrote:
> At 11:05 AM 1/16/2006 -0600, you wrote:
>> Keith Henson wrote:
>>> Corporations already have legal rights, a company can own the
>>> hardware on which the AI is implemented. Corporations can own stock,
>>> so a corporate AI that owns a controlling block of stock in itself is
>>> a free agent with effectively the same rights as meat persons.
>> No, I don't think a computer hardware/software system can own shares
>> in a corporation. Neither can other objects like a vase or rock.
> You missed the point. A corporation can own stock in corporations
> including itself. The corporation can own anything including hardware.

Your original post: "a corporate AI ... owns a controlling block of stock in

That doesn't make any sense as far as I can see. Legally, an AI being considered
akin to a vase or rock, cannot own any shares in any corporation. The term
"corporate AI" as you're using it makes no sense legally.

Can an AI attempt to maintain control over the actions of a corporation via
various methods without actually itself owning the controlling stock? Perhaps,
but as I mentioned in my first message such an arrangement seems rather
pointless and overcomplicated. The AI in such a case is simply property of the
corporation, and relies on its control mechanisms over the humans who actually
perform the actions of the corporation.

>> Generally only humans, and other business entities can legally own
>> shares AFAIK.
>> Also, I don't think a hardware/software system could legally be
>> appointed into any corporate office, such as CEO. At best, you could
>> appoint some humans into those positions whose employment contracts or
>> perhaps the corporate bylaws or something state that they will only
>> take actions suggested to them by the output of the hardware/software
>> system. But if you're forced to arrange things like this, then why
>> bother with all the corporate hassle when in the end you are just
>> property with no legal rights, and relying on a group of people to do
>> what you ask - might as well skip the corporation and just hook up
>> with a single human owner and be their personal property.
> The point of the post I was responding too was how to provide rights for
> AIs. This is a route, a "legal fiction" if you will that could be
> exploited to generate "legal persons" who would have rights within the
> scope of existing legal frameworks.

See above. The corporation that owns the AI property would have rights, but
exercising and/or defending those rights would be delegated to the human
officers in control of the corporation. The AI itself has no direct rights at all.

> If Eliezer is right about the speed with which a seed AI goes way beyond
> humans, the problem will be humans keeping any rights at all rather than
> figuring out how AIs of human level or above should be treated at law.

Of course, and I happen to agree. But I can't resist being pedantic on this
little thread for some reason.

Brian Atkins
Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:55 MDT