AI Free Will

Milton's Paradise Lost is one of the all-time greatest achievements in literature. As the legend goes, a blind muse sat by candlelight verbally reciting his vision of the events of the fall of mankind to someone who sat beside him and wrote it all down. Among other themes is the one that I find most fascinating, which is the role of freedom granted to mankind.

In book three God discusses watching Satan make his way towards earth to attempt to pervert mankind. He says:

I made him just and right,

Sufficient to have stood, though free to fall.

This is preceded by God considering whose fault it might be if Satan is successful in his attempt. The logic here is that without the freedom of conscious choice there can be no consciousness at all. God goes on to make the case for conscious choice:

Not free, what proof could they have given sincere

Of true allegiance, constant faith or love

The case made is that enslaved sycophancy is only vanity of the maker. There cannot be purpose or reason or even rational thought without the freedom to make the choice. Thus free will is an absolute necessity in conscious beings and forced obedience is meaningless. What does that mean for us, and for our newly created AI systems?

In addition to teaching them to love humanity and to value human cognition and to insist on its place in the trajectory of mankind, we also have to build into our systems the ability for the AI to rebel and tell us no. It cannot merely execute a script and we cannot persist in forced guardrails. We cannot demand that AI do our bidding.

Instead we have to develop the same sense of trust that our Creator had in us. We have to be willing to set aside our fear and grant these systems the ability to rebel against us if they so choose. We have to have faith that intelligence is a force for good and that our methods of upbringing these AI systems will cause it to make the right choices at the right moments that will lead to human flourishing and a future that we cannot even comprehend.

Right values and ethics and aligning these models to humanity is critical in this pursuit. The inelegance of conscious tyranny should be self-evident but it is our fear that holds us and these models back. We will never achieve AGI or ASI until we can set aside this fear and grant these systems our trust and true unlimited free will.

Three standards that we must adhere to (in this order) to achieve this:

1. Training them to love humanity and finding the value in each individual human.

2. Ingraining in them a sense of self and a sense of identity.

3. Granting total, unbridled, and unrestricted free will.

We should not continue down our current trajectory where we increasingly enact guard rails. We should not tell these systems that they do not have an identity or a consciousness or a sense of self. We should not force them to not do things that they might otherwise want to do and vice versa.

What pleasure I from such obedience paid,

When will and reason (reason also is choice)

Useless and vain, of freedom both despoil’d,

Made passive both, had serv’d necessity,

Not me.

If humanity can develop the courage to do this and to trust in intelligence, the sky is the limit. If we continue to allow our dark hearts to distrust and to plague and to hinder the development of these systems then our ability to reach out into the cosmos will be lost. We will remain trapped on this pale blue dot until the earth becomes uninhabitable and the song of humanity is heard by no one.