MovieChat Forums > Avalon (2001) Discussion > The Last and Best Theory

The Last and Best Theory

posted 9 years ago by emvan
2 replies | jump to latest

Short version: Ash is a conscious program in a VR world; she's just computer code, but she has free will and a soul as much as we do. The game Avalon is designed to find the "best" such AIs and offer them a chance to be promoted into full, embodied reality. That is the journey to Avalon. (And, yeah, it's at the same time a flat-out religious metaphor.)

All of that can be inferred from the movie. From the above, we can infer that the true world of Avalon is far-future, ultimately utopian, but severely depopulated. It may be on a post-apocalyptic Earth, but it may well be a space colony somewhere in the Pleiades, the star cluster named after the Seven Sisters of Greek mythology. Perhaps physical reproduction doesn't work as well as it ought to. Our descendents have built Ash's reality in order to add worthy souls to theirs.

Ash's reality is based upon an actual dystopian near-future that humanity passed through on the way to this utopian future. The Avalon war game is in most ways a copy of the actual war game that people played to escape this dismal reality. IOW, the opening text exposition is true. The (amazingly smart) catch is that we are not watching that reality, but a simulation of it created much later.

Note that the AIs that populate Ash's reality are being given an opportunity to do just what the actual humans of that time did: escape the reality to find something better.

The "Special Class A" level (known as "Class Real" only to embodied humans) is a VR based on ancient records of turn-of-the-21st-century Warsaw. Why? Why not? Perhaps one of the chief designers of this multi-level VR is of Polish ancestry and possesses and cherishes filmed records of the city. It is, of course, the final testing ground for AIs, to see if they are worthy of promotion to embodied reality.

I'll lay out the evidence and flesh out the created world in the reply.

Prepare your minds for a new scale of physical, scientific values, gentlemen.

[–] emvan 9 years ago

Understanding the Unreturned and Class Real.

First, the phenomenon of becoming "unreturned" is probably not from the original wargame; it's unique to "Avalon," the version in Ash's reality. There's really no good reason why you'd fry your brain playing a VR game, and that risk would have severely cut into its popularity. In Ash's reality, however, programming resources are limited, as evidenced by all the characters who are not full AI programs (and hence which correspond to NPCs in contemporary games). The designers of Ash's reality want to continually weed out the least successful AIs so that they can be replaced by new ones (who would appear to the others as strangers who have moved to their city). So, if you fail badly enough at the game, the gamerunners kill your program; in the VR, your old body becomes one of the unreturned.

Note that we are told that all of the players who are known to have tried Special Class A end up unreturned, but that does not imply that they constitute a majority or even a significant fraction of all of the unreturned. The majority are almost certainly unskilled players who died too many times and/or at too elevated a level (hence the importance of declaring a reset).

Now, an absolutely key clue is the Game Master's question to Ash about the balance between a game that appears winnable but isn't, and one that appears impossible but isn't. This is a clue that it is indeed possible to win at Special Class A.

Why do the players in Ash's reality believe otherwise? Because, of course, everyone who tries ends up unreturned. But they probably include just a very few who failed the ultimate test and whose programs were retired. The real test is having the courage to try Special Class A, and/or the brains to figure out that it must be winnable. It would be stupid of the embodied humans to retire their best and bravest AIs just shy of the finish line.

So what usually happens to the AIs who try Special Class A is that they succeed in their task and their programs / souls are embodied (in a robot or android, one would think). This means that their programs are no longer in their old VR bodies, and hence they join the ranks of the unreturned (whose programs were destroyed rather than embodied).

One bug in this system is the occasional AI who decides that they like Class Real and who chooses not to complete their task. So, sometimes the task assigned to another AI is the apparent killing of that AI. (There are presumably other tasks that those who make it to Class Real are given.)

This tells us what happens to Murphy. He has certainly not proven to the embodied humans that he is worthy of promotion, but he hasn't really flunked the test, either. He was just confused. He has shown some nobility by taking the bullets from his gun. So when Ash shoots him, he is very likely resurrected
in another part of Class Real, and given another chance to understand that he needs to complete his task.

Prepare your minds for a new scale of physical, scientific values, gentlemen.

[–] emvan 9 years ago

The Science of Conscious, Freely Willed AIs.

Ultimately, it's all certainly wrong, but until I finish and publish my book on the topic, it's real science! Folks like Ray Kurzweil believe that if we build a robot AI with a brain that works just like ours, it will have conscious experience, including the sense of free will that we have. (This is the position John Searle has labeled "Strong AI"). And philosophers like Daniel Dennett believe that if we have the sense of free will, and if our behavior is so complex it couldn't be predicted, then we truly have free will, even if our behavior is ultimately determined by the past and the laws of physics (the philosophical school known as "compatibilism").

So a sufficiently complex AI would have (according to Strong AI, which, however, is wrong) consciousness and (according to compatibilism, ditto) free will.

There's no reason why such AIs couldn't be part of a huge VR artificial reality, rather than embodied in robots.

In fact, if all this is true (it's not, but pretend it is), it pretty much follows that we are living in such a VR, as Nick Bostrom has argued. Given the probably near-infinite computing power of our descendents, we're probably living in a simulation they've constructed. But that argument only works if programs can become conscious.

Prepare your minds for a new scale of physical, scientific values, gentlemen.