tawaycozstoned
In the end you want to know what did go wrong (the public will demand it and they are right) and it might be misclassification.
But that is not enough. You want to know why did it classify situation X wrong? So the answer is because the inference network (which was created via the training network) computed its weights because the input was Y. Now you might throw your hands in the air and say "oh it's complicated, the network is undeterministic, blame NVidia", but you can also go farther and build your networks deterministic (which is possible and AFAIK not a performance penalty). Compiler research helps in at least guaranteeing that certain parts of code are deterministic which makes it easier to debug and maybe avoid complex NN misclassification scenarios, but the way to do it doesn't have to do much with NN's itself but more with language design and (real-time, in particular deterministic) OS research.
So the statement oh, that's so complex, we do not no why we misclassified is no excusion, we can do better.
For starters we have to publish NN papers with implementations that describe how to make a particular NN out of given trainigs data (provide that too). We already publish the code and network structure (see caffe, etc.), but often with pre-trained models that have been build on a cluster with many forms of training-data going through network structure, etc.
Now at the moment you read a paper, head to the published code (often the case, again a desirable property of the ML community) and try to reproduce some examples by training on the data.
However it is hard to say in the end, if your network is really as good as the published, as a simple
$ diff my-net-binary-blob.dat tesla-net-binary-blob.dat
might fail, if the stack (training etc.) to build the network is undeterministic. However if you have a good (i.e. deterministic) stack you might be able to reproduce NNs bit-for-bit, which makes it simpler to answer the question "why did we missclassify".But why is that funny (haha) or funny (strange)? If they can build a self-driving car for the masses Elon Musk will be the uber-tech guy for a whole generation, and Tesla seen as one of the good guys with cool tech.
One idea to get correcter code is clean code layout (I heard good things about the LLVM code base), simple to read code and compilers that exploit theoretical knowledge we gained over years of compiler and type theory research. I think Chris Lattner has expertise in all of them (or at least knows about their importance, contributions and drawbacks), and if want to build a full-blown self-driving car it is important to have no indeterminism in your car, and advanced languages and compilers help to guarantee specific statements about your code.
So it absolutely makes sense to invest in your compiler research team for safety, as our security, reliability and correctness expectations will rise (which is good) in particular for self-driving cars (they are not just DNNs).
I can for 100% assure that I only nitpick because I want to understand the logical reasoning for down-voting and are scared that I am too high to understand the meaning of the phrase "It's funny you mention that". I am not a native speaker, but I know all the words, heard it in English and there exists a translation into my language and the interpretation is the same for both, so long story short ... why is it funny?