Hacker Newsnew | comments | show | ask | jobs | submit login
From cosmism to deism: AI makes deities plausible (kurzweilai.net)
22 points by joshrule 1723 days ago | 30 comments

The SAP states that the values of the constants of the laws of physics are so fantastically, improbably finely tuned to allow the existence of matter and life, that it seems highly likely that these values were predesigned.

Even if you accept this (and it’s worth noting that only the tiniest sliver of our universe is hospitable to our kind of intelligent life), it just pushes the question back one level. If our universe was created, then whoever created it must have also grown up in a universe that could spawn intelligent life. Would that universe need to have had its variables tuned as well?

This is just a repackaging of a standard argument that adds complexity while explaining nothing.


> You know, the most amazing thing happened to me tonight. I was coming here, on the way to the lecture, and I came in through the parking lot. And you won't believe what happened. I saw a car with the license plate ARW 357. Can you imagine? Of all the millions of license plates in the state, what was the chance that I would see that particular one tonight? Amazing!

-- Richard Feynman


The best kind of quote- entertaining and instructive.


It's actually much worse than that. Not only would whatever created it have to live in a universe that allowed some form of intelligence - which is unsatisfying, but not all that much worse than "big bangs happen"; not only does it need a creator, it needs a creator with a universe that is computationally far more powerful than our own.

After all, the only way in which we can simulate physical phenomena well is by actually running the experiment. Sure, computer software is very helpful when building a plane, but we still need wind tunnels. I'm convinced that we will create more powerful computers, but I'm not aware of any claims that a particle can be used to accurately simulate two particles... or even one millionth of a particle. (Note that reality doesn't seem to have accumulated any rounding errors even after running for a couple billion of years.)


That sounds a bit like Lee Smolin's idea about natural selection at the cosmological level:



"There are now thousands of AI scientists around the world (concentrated largely in the English-speaking countries) who feel that humanity will be able to build massively intelligent machines this century that will be hugely smarter than human beings."

I'm someone skeptical about this claim, certainly it's a while since I worked in AI research, but from what I recall after the AI Winter researchers were generally pretty guarded about making claims for there being any clear route to building general intelligences.


You don't have to be an AI researcher to give serious consideration to Kurtzweil's statement. We have already built machines that are hugely smarter than human beings, albeit in somewhat limited domains (Google search index is one big example). It is only a matte of time (and economics) before those limited domains become general enough that we will be able to say that, "yes, indeed, the computer so-and-so is generally intelligent." We humans are very good at being humans, so I think it is unfair to judge the intelligence of a machine by how similar it is to us.


That sounds reasonable, I would certainly be quite careful with such statements if I was in that field.

I'm slightly skeptical towards the opening part, are there really "thousands of AI scientists" in the English-speaking parts of the world? I'm surprised, but I have really no insight into the world of AI science.


Though Kurzweil is well respected I find myself unconvinced of most of his assertions.


Your not the only one - http://scienceblogs.com/pharyngula/2010/08/ray_kurzweil_does.... I personally think he's completely ridiculous.


"Theism is the belief in a deity that also cares about the welfare of individual humans. Deism I am open to, whereas I find theism ridiculous. The evidence against it is enormous. For example, last century, about 200-300 million people were killed for “political reasons,” e.g., wars, genocides, purges, ethnic cleansings, etc. It was the bloodiest century in history."

Unless that creator with a brain that makes yours look like that of a mouse's cares about the welfare of you. Made sure all those wars and killing occurred in just the right way just so that you could be here writing this article. If your brain is just a tiny mouse brain how could you possibly know?


No. AI only makes deities plausible if you subscribe to a very narrow and particular Singularitarian school of thought on what AI is and where it's heading. The megaphone diplomacy of these people tends to detract from the credibility of AI research in general, by associating it with very unrealistic views and fallacious thought experiments which as far as I'm aware originate from the opinions of Terrance McKenna.



Regarding Kurtzweil's "mathematical principle," our mathematics is a reflection of the universe we live in. Mathematics is based on axioms, which make sense to us only because we observe them to be always true in our universe. For example, one of Peano's axioms of arithmetic states that "1 is a natural number" (here I am using the base one axiomatic), and another axiom states that "no natural number has 1 as its successor." One could imagine (though not necessarily describe in detail) a strange universe where there is a natural number larger than 1 that has 1 as its successor (i.e. the way modular arithmetic works).

To summarize, I do not see why the "mathematical principle" has to be postulated separately from the "anthropic principle," which we all know has an alternative deity-free explanation mentioned by Kurtzweil himself: there could be "a zillion universes, each with a different set of physical laws, and we just happen to live in one that is compatible with life."


Regarding Kurtzweil's "mathematical principle," our mathematics is a reflection of the universe we live in. Mathematics is based on axioms, which make sense to us only because we observe them to be always true in our universe.

How about: Our mathematics is a reflection of the universe we can perceive.


That is correct, except that it puts unnecessary emphasis on perception relativism. We use mathematics especially extensively in areas where our perception is limited (particle physics, climate modeling etc.) In other words, we use mathematics to "augment" and to "test" our perception so as to prevent any kind of perceptual illusions misguiding us in our observation.


That is correct, except that it puts unnecessary emphasis on perception relativism.

Well, in your mind. :) I didn't mean that in support of a relativistic viewpoint. Perhaps I should have said that we can comprehend. Perception just determines our starting point.


"Personally, I think if science could come to the conclusion that there is/was a deity that created the universe, then that would be wonderful for science. It would open up a vast new arena for science to play in."

No, it would not. If science could "play in that arena," science will have solved the halting problem.

And have defined an algorithm for determining any possible busy beaver number. Which, of course, is related to the previous.


"The rise of artilects (artificial intellects, i.e., godlike massively intelligent machines with intellectual capacities trillions of trillions of times above the human level) in this century makes the existence of a deity (a massively intelligent entity capable of creating a universe) seem much more plausible."

Um...where exactly are these artilects? Haven't heard of any recently. Pretty sure we would have noticed. Unless of course he's saying that they will arise in the next few decades, in which case - more of the same old gee-whiz-the-singularity wankage.


I think he's saying that they will rise over the next 100 years (well, 89 actually). Not completely unreasonable if you think of the advances in technology over the last 100 years.


Sure, if you assume that everything will continue on the same trajectory in perpetuity. This is a fairly large assumption with little evidence to back it up.


Here are some temporary mirrors, the main site is slow to respond:




Ok, from the talk,

21 Century A.I.-enabling technologies

a) Moore's Law

b) 1bit/atom by 2020

c) Femto-Second Switching

d) Reversible Computing

e) Nanotech

f) Artyficial Embryology

g) (Topological) Quantum Computing

h) Nanotech Impact on Brain Science

j) Artificial Brains


Mini-universes or not, the first society on Earth to create and to use artilects – even ones only a magnitude intelligent than humans – will use strategic superiority to expand and to dominate most parts of the earth, in much the same way that Western society came to dominate North and South America and Australia during the Age of Imperialism.


"as humans, with our puny human brains, trying to imagine what an artilect would think about is like a mouse trying to imagine what humans think about, using its puny mouse brain.... Deism I am open to, whereas I find theism ridiculous."

There's a disconnect there, and I've seen it for a long time. The reason theism is called ridiculous is that the author imagines what their millions/billions/trillions/more-times intelligent entity would do in our universe, looks out and observes this is not what is happening, then declares that there is no Theity (to coin a word). But there's a major fault in the logic, which is that one can be so confident about what that entity would do that the imagination step has any meaning. Researchers like Kurzweil who otherwise would never dream of claiming they can read the Theity's mind will, on this one point, declare that they can.

I can't prove the conclusion false any more than the next guy, but I can say this argument is fallacious.

I also remind you I'm not defending any particular Theity; consider the Riverworld scenario, for instance: http://en.wikipedia.org/wiki/Riverworld Perhaps the Theity only cares about the eventual creation of achilects and is indeed interested in us, but only to the extent that we eventually produce one. Maybe archilects inevitably resurrect [1] their creator races and the pain and suffering to get to this point is but a passing dream, and perhaps that pain and suffering itself has further impacts on the ultimate form of the archilect in some important way (a being with no concept of the perils of "evil", for lack of a better word, should probably not be trusted to build Utopia). Perhaps there's a theity that is interested in silicon-based life and doesn't care about us, but still cares about other individuals in exactly the way that Kurzweil declares impossible. Any number of possible scenarios. My point is simply that you do not have exceptional insight into these being's minds on this particular point; you are uniformly ignorant about them.

I really don't know what the truth is, even as I've put my markers down, but I am increasingly sure that it is weirder than any of us really can imagine. I only hope it is for the better.

[1]: Which in this context isn't mystical, either; "resurrection" is "merely" somehow obtaining a copy of the mental state of a being with sufficient detail to once again begin simulating or embodying that being again. If we're seriously talking creating universes, there's a number of possible avenues for this to take; an as-yet unknown way of reading the past state of the universe directly, simulation of the race up to that point with this level of detail (which need not be atomically accurate, just human-scale accurate), or most likely some approach we can't even imagine.


I'm inferring a distinction that Kurzweil doesn't make explicit, but I'd guess he regards "theism" as a claim that the creator possesses specific (highly anthopomorphic) traits, eg. love, or personal interest in our lives. I doubt he would recognize much difference between the unknowable theity that you describe and his own "artilect." (quoted because I think it's a stupid word). For theists, the concept of God necessarily carries the baggage of what God is like.

The Problem of Evil, philosophically, has never been a refutation of the existence of God. It's a refutation of the existence of a particular kind of God.

As an aside, one of the most common answers to the Problem of Evil is that God's justice is beyond human comprehension. I think it's interesting that this dovetails the debate of philosophy into semantics. That is, if God is just, but not in a way recognizable to humans, is it even semantically meaningful to call Him just?


I think your description probably does match the actual baggage of the term in his mind, but he in fact did make something explicit, right in the previous sentence to one of the two I quoted: "Theism is the belief in a deity that also cares about the welfare of individual humans. "

You do not and can not know how what I was calling a Theity cares. It is itself a very primitive idea that the caring must manifest as heavyweight, visibly-obvious intervention into everyday life. I do not deny that this is certainly how many or most conceive of it but it should be pointed out that not all religions that contain Theities have this idea actually deeply embedded into them (as opposed to deeply embedded into the minds of the bulk of practitioners), but actually can still work with a relatively subtle Theity using the power of incomprehensibly vast intelligence to accomplish their goals through the smallest of possible changes, an elegant approach.

(Actually Kurzweil's characterization of religion is rather paper-thin, too: "Presumably, millions of those killed were theists, believing that their “theity” would “look out” for their welfare." I know that's not intended as an explanation of The Totality of Religion in one sentence, but I still shouldn't even have to describe how this does not match up with the contents of many religions, particularly including Christianity which promises persecution, explicitly, several times. I think many religions and many actual expressions of religious are rather more subtle than he is willing to give credit for.)


You do not and can not know how what I was calling a Theity cares. It is itself a very primitive idea that the caring must manifest as heavyweight, visibly-obvious intervention into everyday life.

It turns out my aside does address this. If a theity is said to "care about the welfare of individual humans", the caring must actually be something that humans would actually recognize as caring. Otherwise, you're using the word to connote something that it doesn't, in order to piggyback on the emotional resonance of what it actually does.

Saying God may be caring but we don't know what that looks like is exactly equivalent to saying we don't know anything about God. The entire point of theism is to claim that you do know something about God.


Ah, in this case I simply mean caring in the weak sense of "interested in", which I think is the most generous way to read Kurzweil's point. If you really get technical, no words have any particular meaning about any entity of this intelligence, but it would be at least reasonable to define some sense of "interested in" us as more than just an incidental detail of running a universe. From what I've seen Deists-but-not-Theists tend to mean an entity that literally does not care about us, may not even know we exist, would not care to know about the ignorance, and sometimes I even get a whiff of the self-loathing-human sense of "and if it did know we existed it wouldn't like it". Again, not necessarily always spelled out but I think it's pretty much as fair as your previous assessment of Theism, which is to say, a decent working definition of what is already out there.

Your side point I actually deliberately left addressed because it goes down a rabbit hole. A fun one, but not one I was trying to go down. :)

It is not a new idea in theology that we don't really know the mind of God and that the words and concepts we use are merely approximations. It does rapidly deteriorate into something unprovable. Unfortunately, where in science we can discard the unprovable and justifiably consider ourselves wise, I think the undisprovability and/or unprovability of all these questions is actually fundamental and unavoidable, inasmuch that observing that certain statements are unprovable doesn't affect their truth value.

You could define a concept of "human justice as the considered timespan approaches infinity", for instance, which may still be vague but is at least getting somewhere, sort of.


You might compare author Hugo de Garis's position on deism with the Simulation Argument - i.e. our present projections of technological progress shape the expected chance that we live within a reality manufactured by beings very much like our own descendants.


de Garis' works are well worth reading, though like many AI-focused futurists - such as those associated with the Singularity Insitute - I think he gives too little weight to comparative advantage when projecting future interactions between beings of vastly different capabilities.


You might give it too much weight--remember, comparitive advantage says humans have something to offer AIs; it doesn't say whether that something is worth more than our care and feeding or constituent atoms.

It's easy for the income level guaranteed by comparitive advantage to fall below subsistence income; it happened to most horses a century ago.


This is definitely going to help with the accusations of religiosity and crackpottery.


Applications are open for YC Winter 2016

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact