The biggest challenge in creating safe AI, is called the “Alignment Problem,” — aligning the AI’s goals with humanity’s goals.
To solve this problem, experts recognize that it will require universal agreement on core human principles.
Without pre-defined moral absolutes, history has proven the results will be disastrous.
And, unlike overthrowing a tyrannical government, there may be no way to put the genie back in the bottle once super-intelligent machines are released with ill-intentioned puppet masters controlling them.
You can’t accept that all human life is valuable and also support eugenics, racism, or slavery; so, AI needs to be programmed to value all human life.
The reality in a nutshell is that nearly all value systems can result in significant harm to humanity. Without core human principles which are also called moral absolutes that value all human life, there is a significant risk to humanity.
Why are you and me Infinitely Valuable.
Once you clearly answer this question, you can then train the AI to respect human life without exception. However, if you can’t clearly define why human life is intrinsically valuable, AI will also have no reason to prioritize human life.
Many philosophers have posed various things that make humans unique including: Intelligence, Consciousness, Moral Reasoning, and Curiosity.
However, all of these will ultimately fail because there’s no way to measure these things in absolute terms so AI, and many people, may ultimately conclude that AI is superior to us in all of those things.
Defining what makes a human unique must be a binary measurement.
Binary is like a switching on the kettle; it’s either off or on.
There are no degrees of humanity. You are either fully human or fully not.
So, what is a binary definition of a human?
They have immortal souls.
For super-intelligent machines to value human life beyond anything else, they too must be trained that God is the source of our unique value in the universe.
Quite clearly, without our God-given nature, we value our own species above others simply as a matter of preference.
Other attributes that we might consider that separates humanity – creativity, emotions, feelings, empathy, abstract thought, symbolic thinking, being alive, consciousness, advanced social interactions, curiosity, moral reasoning and others – may be unique.
But these aren’t binary as they all exist on a continuum (i.e. some people are better at abstract thought than others).
Therefore, using the above attributes to define what makes a human unique can lead AI down the path to valuing people according to their capabilities instead of valuing every life absolutely.
The reality is training AI based on consciousness, intelligence, moral reasoning, or sentience is, first, impossible to quantify and so ripe for exploitation by bad actors and, second, only measurable on a continuum where more intelligent machines would then be more valuable than humans.
How To Solve The Alignment Problem.
In simple terms, for the AI to align with humanity’s best interests, we first need to define universal human values (moral absolutes) for the AI not to violate.
To train an AI (Large Language Model) that certain things, like murder, rape, and slavery are wrong requires three things. Belief in God Moral Absolutes
Humans are uniquely made in the image and likeness of God. To add more secular terms, you must train the model that human life is valuable and uniquely so in a way that can never be eliminated. You have to teach it to respect the sanctity of life.
Human life is uniquely special, because we are made in the image and likeness of God. To be clearer, all life – including our immortal, immaterial souls – are a gift from God.
The only way to prevent potential destruction of humanity is to require AI to be trained on the same things civilized societies have taught for millennia.
God exists.
Moral Absolutes – primarily to protect the sanctity of all human life. Our immortal soul is the source of the dignity of human life and therefore no human, society, government, opinion, scientific study, alien, or evolving machine can diminish that dignity.
This is how you solve the alignment problem.
Yes, there are bad people who will now have access to the most powerful tool for propaganda we’ve ever seen. If we don’t make it illegal to train AI in ways that don’t respect all human life, the consequences can be disastrous.
AI only knows what it’s trained on as AI will never wake up and become conscious. If people with bad intentions train it with immoral values, it will follow those values without question.
That’s all for now folks.
Thanks you so much for reading this 5 Part AI series.
In October 2024, I will be offering products, memberships and services for those who want to discover their mission or improve the one they already in the age of AI with my new and unique process.
If this is something that resonates with you welcome to the “New Resistance”. I have been waiting for you.
Together lets build better lives and worlds for ourselves and those we love and for all other “sovereign humans”, on this wonderful planet.
I welcome you to continue your journey with me and get onto my email list to be notified when my new products and services go live.
With Love and Dharma Always,
Manoj