This is true, but the comparison isn't relevant. We find it cheaper to manufacture systems that aren't fault tolerant in most cases though we do no how. It is an economic decision. Therefore, in practice computer code is FAR more fragile than DNA which has numerous redundancies and fault tolerance mechanisms built in (once you boot up an organism, you can't shut it down or -HUP the process). The bottom line is that it serves living organisms well to have a code that functions well when subject to a high error rate. Life does not need deterministic results to function. If something non-deterministic happens in a computer system or an accidental code mutation occurs, we WANT the system to crash. The onus is on computers to give the right answer every time or they are useless for our purposes.
That's a total non-sequitor. It doesn't even matter whether or not DNA code or human programming is more fragile, as neither answer is germane to this discussion. What matters is whether useful programming can happen without any form of Intelligent Intervention (i.e. a simulated lifeless, primal, unintelligent environment).
For human programming, we know for a fact that useful programs will never form on their own in a computer, no matter how much static or noise is in the computer environment. For DNA coding, we know for a fact that we've failed to get useful DNA mutations in natural, unaided lab environments (or in the wild, for that matter).
Thus, the only tangible evidence which we have in hand are human programs formed via intelligent intervention rather than through natural, unaided events. One can speak of "possibilities" all day long, but science requires both evidence as well as repeatability. Clearly we can repeat ad nausium the intelligent creation of useful human programs, but the scientific jury is still "out" on wether useful programs can self-form without intelligent intervention.