Replies

It’s become popular for the developers of these bots to embed within them a “constitution” or a “soul” document. Here is an article on Claude’s 23,000-word document: https://youmind.com/blog/claude-constitution-ai-alignment-guide

There seems to be a view in the tech bro space that all we need is more data or rules or documentation etc., and the bots can achieve human-level cognition. To be fair, models and neural networks and chatbots have come quite a ways. They CAN do some interesting things.

However, they will never be sentient, or anything close to that. Why? Because every soul has moments - more precisely split-seconds - where decisions are made and we later ask “where did THAT come from?”

Maybe THAT was based on years of ethics, experience, or preferred patterns. Maybe in those high-pressure moments, the brain acts in a way that can’t be predicted…and may not be repeatable.

Or, if we are honest, there is the Holy Trinity that works in ways that NOBODY understands in those moments.

Think of the times when a ball bounced in front of your car and you swerved left and missed a child; if you’d swerved left you’d have killed little Mary. In addition, ‘Something” told you to take a different road and you avoided an accident. And then there was that time you give a killer answer on an interview to a question from left field that you couldn’t have worked out that spectacularly with preparation…and you got you that job. You met your spouse due to a chain of events that were impossible, improbable, and baffling.

There is a saying, man plans and God laughs. Perhaps we can update it as God laughs at Techbros and PowerPoint.

It’s become popular for the developers of these bots to embed within them a “constitution” or a “soul” document.

RoboCop is programmed with four prime directives by his corporate creator, Omni Consumer Products (OCP):

Serve the public trust
Protect the innocent
Uphold the law
Any attempt to arrest a senior officer of OCP results in shutdown (initially classified)

This is Penrose's point. We can't even explain human consciousness, don't know how it arises; so how do we build it into a machine? He says that human understanding is not computational and involves more than we can map, more than mere brain networks, neuronal activity, etc. There's something going on that we can't pinpoint or even name, and it's an ultimate sine qua non to the whole matter.

Again, AI may be very useful in many applications. It will never be sentient or conscious and will never be able to do all that the human mind can do in any application.