Entity agendas & rights (was Re: Robots in Social Positions)

Robert J. Bradbury (bradbury@aeiveos.com)
Thu, 8 Jul 1999 22:01 PDT

> "Mark Phillips" <clay8@hotmail.com> wrote:
> "entities," or even [sometimes] "robots"). And, indeed, these systems,
> robots, what-have-you, well be not only intelligent, but ULTRA-intelligent.
> WIll they qualify as *persons* (a la Strawson, David L. Norton, et al)?
> Tough question. My intuition is PROBABLY!!

I'm unfamiliar with the authors you mention, but there is a very good discussion of this in Stanislaw Lem's "Golem" story in "Imaginary Magnitude". It points out how we will not even be able to begin to grasp (or relate to) an "ultra-intelligent" machine.

I think "person" is a poor term. The first question I would ask is -- "is it self aware)?" (going back to my "mirror-test" discussion of a few days ago). The second question I would ask is "does it have a self-"agenda"? (i.e. a will to survive or goals). Now in humans, nature has built-in the survival and reproduction goals (required for the system of natural selection and evolution), and presumably these play in important role in the management of "consciousness" (which seems to determine ones "agenda"). In Robots (at least the non-self-evolved ones), I would envision that we program-in the goals/agenda. [So assuming the Robot can pass the Turing test (or your personal test for human-class intelligence), *do* we split-hairs on the difference between an unconscious-nature-derived self-agenda and and a human-programmed (seeded?) self-agenda. [I.e. you aren't a real "person" unless you have a "natural" agenda.]

Now, assuming some humans rise above the natural agenda and write their own agendas, is then that the criteria for "high-level" persons (akin to high-level languages perhaps?). So now you have 2 classes of both humans and robots, those that write their own script and those that don't.

An entirely separate question is whether or not the human/robot deserves "rights". This is an entirely legal/political concept. Historically, slaves and women have suffered from a lack of "personhood" (and therefore a lack of certain rights). I think some people have a problem with the idea that "machines" might have human-equivalent intelligence, but many more start to voice objections when it comes to granting them human-equivalent rights. Thats when things get really loud, "After all, they are only machines, they aren't even *alive*"!

Lem's Golem, if I recall escaped from the whole problem by self-evolving into a higher dimension. [The Golem explained our shortcomings to us, then left us to stew in our own juices.]

> And don't forget that "we" shall soon-on be able to *merge* with
> entities/systems such as these. What of personal identity??

I think the current thinking is that each copy of something with its own history, i.e. in programming terms each "instance" would have its own identity. "Permutation City" is a great book that examines this to some degree.

Now, at the Foresight Institute Senior Associates meeting this year they had a *very* interesting exercise related to this.

Say one Ralph Merkle (if you know Ralph, this scenario becomes very funny), gets himself voted into power in a country where the leaders have a lot of power, lets say Brazil. He then passes a law that says "uploads and copies of uploads" have the complete rights of humans (this makes complete sense, yes?). He then has himself uploaded and diverts some of the resources at his disposal (after all he is the president and magtape is cheap) to making 93 million copies of his upload. Now to avoid having to support all of these uploads, distorting the economy, etc, etc. these uploads are all in "suspend mode". However, the day before the election he "activates" all of his uploads. Well of course, these uploads think very highly of Ralph and re-elect him president. Since their prospects for work aren't very good, they simply decide to resuspend themselves until the next election comes around.

This is a sticky little problem. It creates havoc with the concept of one "person" = one vote. Now I'm sure you can find flaws with the scenario as I've presented it. Many of these were dealt with in the discussion to make the situation quite plausible. The real issue is - how do you distribute, grant, allocate, etc. "rights" when you have "entities" that can be created cheaply and may have "programmable" agendas (to greater or lesser degrees)... You can't easily say that a copy of a person is not a person (at least from my frame of reference)!