An alert reader pointed me at a recent post over at Uncommon Descent by a guy who calls
himself "niwrad", which argues (among other things) that life is
non-computable. In fact, it basically tries to use computability
as the basis of Yet Another Sloppy ID Argument (TM).
As you might expect, it's garbage. But it's garbage that's right
up my alley!
It's not an easy post to summarize, because frankly, it's
pretty incoherent. As you'll see when we starting looking
at the sections, niwrad contradicts himself freely, without seeming
to even notice it, much less realize that it's actually a problem
when your argument is self-contradictory!
To make sense out of it, the easiest thing to do is to put it into the
context of the basic ID arguments. Bill Dembski created a concept called
"specified complexity" or "complex specified information". I'll get to the
definition of that in a moment; but the point of CSI is that according to
IDists, only an intelligent agent can create CSI. If a mechanical process
appears to create CSI, that's because the CSI was actually created by an
intelligent agent, and embedded in the mechanical process. What our new friend
niwrad does is create a variant of that: instead of just saying "nothing but
an intelligent agent can create CSI", he says "CSI is uncomputable, therefore
nothing but an intelligent agent can create it" - that it, he's just injecting
computability into the argument in a totally arbitrary way.
So, what's CSI? Long-time readers will have seen my old
critique of it. I'll just reiterate the key points here. CSI is something
that you can never really pin down: it's a contradiction wrapped up in
obfuscatory mathematics to make it appear meaningful. Nothing
actually has specified complexity, because nothing can have specified
complexity, because specified complexity is fundamentally self-contradictory:
by looking at the basic definitions of the terms using information theory, you
find that specification equals not-complex, and complex equals not-specified.
So to have specified complexity is something like being both invisible and
florescent pink at the same time.
Ok, background out of the way. Let's look at his article. In the first
section, he presents his version the standard CSI argument, with the random
insertion of computability. I think the best summary of it is the following:
IDT shows that CSI cannot be generated by chance and necessity
(randomness and laws). An algorithm (which is a generalization of law) can
output only what is computable and CSI is not. The concept of intelligence as
"generator of CSI" can be generalized as "generator of what is incomputable".
Obviously, needless to say, intelligence eventually can generate also what is
computable (in fact what can do more can do less). Intelligence can work as a
machine but a machine cannot work as intelligence. Between the two there is a
non invertible relation. This is the reason why intelligence designs machines
and the inverse is impossible. To consider intelligence as "generator of what
is incomputable" makes sense because we know that intelligence is able for
instance to develop math. Metamathematics (Gödel theorems) states that math is
in general incomputable. It establishes limits to the mechanistic deducibility
but doesn't establish limits to the intelligence and creativity of
Now it's straightforward to see that the generator of what is incomputable
is incomputable. Let's hypothesize that it is computable, i.e. can be
generated by a TM. If this TM can generate it and in turn it can generate what
is incomputable then, given that an output of an output is an output, this TM
could compute what is incomputable and this is a contradiction. Since we get a
contradiction the premise is untrue, then intelligence is not computable.
Before I get to the meat of it... Gödel's theorems don't say that
"Math in general is uncomputable". I'm going to pick on this, because
I've mentioned Gödel many times on this blog, and I've frequently
been guilty of over-simplifying when I talk about what Gödel's incompleteness
theorem actually says. It's hard to state simply in a way that actually gets
the true depth and meaning of it across clearly. But as bad as I've been,
I've never come close to botching Gödel this badly. I don't know
whether niwrad has ever actually studied Gödel or not; I suspect
not, and that this is just his wretched misstatement of his own misunderstanding
of an over-simplified statement of Gödel by someone like me. But it
does point out the danger of having people like me try to present
simplified explanations of complicated things: there are always bozos
who believe that by hearing a simplified intuitive explanation of something,
that they've understood the whole thing, and will then go off and run
(What Gödel actually said is something closer to "Any sufficiently
powerful formal reasoning system will be either incomplete or inconsistent. If
it's incomplete, that means that it will be capable of expressing true
statements which are not provable in the system. If it's inconsistent, it will
be capable of expressing statements which are neither true nor
false." And that is, itself, a wretched over-simplification, and I'm willing
to bet that a couple of commenters will call me on it. Trying to state
Gödel simply is really difficult, because it's simultaneously
a simple statement mathematically, while also being incredibly deep
and profound. It just doesn't render well into english.)
But that's not close to the worst part this babble. That's
the second paragraph quoted above: "The generator of what is incomputable
The fundamental example, the first example, the most canonical example of
un-computability that anyone who studies computation knows about is called
the halting problem. The whole point of the halting problem
is that you can easily create a program which generates non-computable
results! The proof of the halting problem shows a completely mechanical
computable mechanism by which any supposed halting oracle can be
defeated - thus showing that the halting problem is uncomputable.
That's also part of what Gödel did. He showed a mechanical
process by which you can trick any sufficiently powerful formal system into
producing problematical statements. You don't even need to understand the
system. I can write a program which takes a description of a formal
system as input, and generates the series of steps to produce a Gödel
statement for that system. So the claim that Gödel proved that
you can't generate uncomputable things by a computable mechanism is
complete nonsense - in fact, it's the opposite of what Gödel
But you can basically take this section, and reduce it to a simple circle:
Intelligence is the ability to generate CSI. Why can intelligent things
generate CSI? Because the definition of intelligence is the ability to
generate CSI, therefore if something is intelligent, it can generate CSI.
Really, computability is just a red-herring: he's equated CSI with
a particular kind of non-computability, and then written the CSI circular
argument with "CSI" replaced with "CSI equivalent noncomputability".
In the next section, he tries to go a bit farther, and not just prove that
intelligence is non-computable, but that the non-computability of intelligence
proves that there must be a God, which is the "infinite information
source.". In this section, he proceeds to disprove his argument
from the previous section.
The argument: Intelligent beings produce information. Information can't
come from nowhere. Where did it come from? In the last section, he said that
intelligent beings can create CSI. But now he's actually reneging on that. The
information that intelligent beings produce can't just come from the
intelligent beings: that would be creating something from nothing, which is
impossible. So there must be a higher being - an infinite information
source which produced all of the information. Intelligent beings
don't create information. They just regurgitate it.
He doesn't seem to have the slightest clue that he's contradicting himself
here. But by the argument in this section, a human being is no different from
a Turing machine: both cannot, by his argument, produce CSI/CSI-equivalent
noncomputable information, unless they've obtained it from some other source.
Suddenly, life isn't "non-computable" anymore - what he describes now is life
as a computable device which takes inputs from the "infinite information
source". (Which is, of course, just another shallow renaming: just as he
substitutes "CSI-equivalent noncomputability" for "CSI", he substitutes
"infinite information source" for "intelligent designer"). It's the
same stupid trick.
After lots of rambling around that basic argument, he moves on to his next
section, in which he proceeds to pretend that he didn't just obliterate his
own argument. He returns back to the argument that since supposedly intelligent
beings can create information, they must be non-materialistic - because
materialistic things can't do non-computable stuff.
The whole computability argument comes down to this little bit of
circularity. niwrad asserts that life in general, and intelligence in
particular, must contain CSI. But he can't prove it. He can't point at
any specific property and prove that it has CSI. Instead he just relies on
intuition: it's just obvious that these things have specified
Then he asserts that CSI can't be generated by any non-intelligent
process. Again, he doesn't prove it; he doesn't even really argue
it. He just blindly asserts it.
Finally, he takes his two assertions: life has CSI; CSI can't be produced
by a non-intelligent process, and based on those, he can conclude that life is
non-computable. Since a computing device isn't intelligent, it can't produce
CSI: CSI is, by definition, non-computable. Therefore, if life (or intelligent
life) contains CSI, then by definition, life is non-computable. QED.
Same old, same old.