A nanobot defence system should work in principle.

Apparently a lot of people seem to overreact to the possible 
dangers of self-replicating nanobots.

However a nanobot defence system functioning similarly to the 
human immune system in principle should work fine. Why is that?

First of all, accidental self-replication and mutation of 
nanorobots can be very easily prevented. 

a) Nanobots will function like current computer systems. 
This means that single point damage will lead to failure 
of the complete system. Of course nanobots with redundant 
systems can be designed. But they will be more complicated 
and less efficient. And, of course, it's more sensible to 
design nanobots especially with single point failure in mind.

Furthermore the internal software can be easily protected against 
mutation by using error detection and correction algorithms.


b) Almost all nanobots will be designed to function only in 
specific artificial environments. (Like cars that can run only 
on refined fuel. And unlike biological systems.) They will need 
very specific fuel molecules and very specific (and refined) 
foodstock molecules deliverd in a very specific environment. 
And again, of course nanorobots that can function, survive and 
replicate in the real world environment could be designed 
(or developed). But they will be (very) difficult to design. 
And they will be less efficient than more simple (limited purpose) 
nanobots. It's even likely that such free environment nanobots 
will be less efficient than biological replicators.

Therefore it looks quite unlikely that accidental uncontrolled 
free replication of nanobots will happen. 


2) Against intentional use of replicating nanobots for harmful 
purposes, nanobot defence systems can be used.

As stated before nanobots designed for a single purpose will 
be more efficient than multipurpose bots. For example a nanobot 
that can only attack and destroy another nanobot, is more efficient 
(smaller, uses less energy, less building resources, etc.) than 
a nanobot that can do so and also selfreplicate. It is more efficient 
to do the production of the defence nanobots using a dedicated 
production system. And to do the selfrepliction using systems 
specifically designed for self-replication. 

Furthermore if two nanobot systems containing bots with equal 
efficiency will engage in "battle". The one containing the largest 
number of bots will prevail. (Assuming both are using equally 
efficient strategies.) So an obvious precaution against nanobot 
attackers is to keep a large stockpile of nanobots on standby.

Of course, this is all simplified. But in principle this should 
hold.

Some people argue that an attacker could infect the nanobots of the 
the defence system (like computer virusses do). But this can be 
counteracted by continuously checking the integrity of all defence 
nanobots (for example by dedicated other nanobots). When an infected 
nanabot is detected (or an intruder) it will be attacked. (Kind of 
like what happens in an ant community.)


So it's not that obvious that a nano defence system will not work, 
or that accidental harmful selfreplication of nanobots will occur. 
In fact, after some thought, the reverse seems more obvious.

-----------------

Some further misconceptions I noticed :

Somebody argued that molecular nanotechnology will not change the 
quality of medicine. But indeed it will do so!

Molecular nanosystems-based manufacturing will provide new tools for 
medicine, making possible molecular-scale surgery to repair and 
rearrange cells. Since disease is the result of physical disorder 
--of misarranged molecules and cells--medicine at this level should 
in principle be able to cure *ANY* physical disease. 


Another person assumed that nanobots will be intelligent. 

They will not be intelligent like humans (or higher animals), but 
more like insects. A standard nanobot will contain the processing 
power of a very simple computing system (like a current process 
controller system, i.e. a simple 8 or 16 bit CPU).

To make a really intelligent system with a lot of processing power, 
such a system will need to be larger than one microscopic nanobot.