Saturday 16 June 2018

Andres Gomez Emilsson's pure replicators

This post is about Andres Gomez Emilsson's proposed concept of "pure replicators". First, I'll let Andres introduce the concept:

I will define a pure replicator, in the context of agents and minds, to be an intelligence that is indifferent towards the valence of its conscious states and those of others. A pure replicator invests all of its energy and resources into surviving and reproducing, even at the cost of continuous suffering to themselves or others. Its main evolutionary advantage is that it does not need to spend any resources making the world a better place.
Conventionally, a "replicator" is something which copies are made of. Agents and minds aren't really replicators, they are large complicated things which can't easily be copied. "Reproducer" might be more appropriate terminology from this perspective.

Evolution optimizes for survival and reproduction. However, that does not mean that it builds creatures that are uniformly devoid of compassion. The evolutionary function of compassion may not be obvious or easy to explain, but it is likely to exist because compassion is widespread among humans. Probably, compassion promotes social cohesion and encourages acts of reciprocal altruism.

Andres warns against becoming a "pure replicator", but he defines this as an agent indifferent towards suffering, and most humans care act as though they care about the suffering of themselves or others - because compassion is built into them by evolution as a proximate goal. Becoming free of compassion does not seem as though it is a likely fate in the first place.

Andres apparently agrees, writing: "Most animals do indeed care a great deal about the valence of their own consciousness". He goes on to explain that "pure replicators" are mostly a future threat. There follows a bunch of speculation about how future intelligent machines might not use the pleasure-pain axis in their motivational systems.

I think that part of the problem here is a failure to properly distinguish between proximate and ultimate goals. A hypothetical agent with the sole goal of maximizing the number of their great grandchildren might still have proximate goals of minimizing the suffering of themselves and others. Many kinds of personal suffering are likely to be negatively correlated with the number of great grandchildren produced. Concern for others could well be adaptive too, though that's a bit harder to understand. That is how evolution can build compassionate creatures.

Andres apparently thinks that compassion is a useless spandrel. That seems tremendously unlikely to me.

However, my number problem is not with the science, it is with the terminology. You can't just hijack the "replicator" terminology and load it up with qualia for no good reason. The proposed termiology is simply ridiculous.

No comments:

Post a Comment