This is actually a problem that I'm working on in my PhD on Machine Awareness so I'll try to give as simple and concise answer to this as I can.
Searle's argument in this case is valid; humans INFER liveness and consciousness on things as a result of our anthropological evolution. This is why children play with dolls and we give our cars (and especially our GPS units) names when they talk to us. It's also why we buy our meat in plastic wrap and get so upset if we visit an abattoir. This is an important point; the sense of liveness is our inference, not a direct implication made by the AI or similar system.
The problem with the Chinese Rooms thought experiment is that language does tend to follow some formal rules. What the Chinese Rooms thought experiment doesn't address is how would you get such a system to answer the question 'What's your opinion on X?' In other words, it's great when the response is canned, but it wouldn't let you express yourself in the Chinese language.
The lightbulb analogy. Hmmm. Let me start by saying that what makes this argument flawed is the idea that intelligence is judged by the output rather than by the thought processes that led to it.
Here's a thought experiment for you; Man walks into a cafe and urinates on the sandwiches. He's arrested and asked to explain himself.
Explanation 1 - 'Well, bread can absorb moisture so it would result in less splashing and I really needed to go'
Explanation 2 - 'Well, I'm trying to make a political statement about the futility of applying a common framework of rules over the top of a society when it stifles true innovation and lateral thinking'
Both generate the same output, but for very different reasons. Which one is more intelligent? Person 1 hasn't factored in consequence, Person 2 has. Person 2 (on the other hand) had no thought of the needs of others, whereas it could be argued that Person 1 did. There's no simple answer to this by the way; what I'm trying to point out is that considering the answer is completely the wrong way to evaluate intelligence because this is largely a factor of the context taken into consideration by the actor.
Another way of thinking about this; in computer science, I can write a really efficient program to get result B from input A, whatever that may be. It might so happen that this really complicated AI also gets result B from input A, but the difference is that I wrote the first program to process data. The second program was written to 'solve a problem'. The only difference between the two programs is the intent of the programmer. That's really important because again (and this part is probably a topic for another time), it is not the computer that implies any meaning to its output. We infer it.
Consequently, the lightbulb analogy doesn't look at the right aspect of the problem. Just because two devices both produce light it doesn't mean they are the same in any way.