[Butler911 (OP)]

Is it possible to upload all our knowledge of science onto a computer that will logically work out the questions we ask it?

[Nizzle]

No, not at the moment.

A lot of discoveries in science still depend on creativity. A computer cannot be creative, as long is it is not intelligent. So first, we need to create human-like, or better-than-human AI. Then we can talk about a computer doing experiments.

[damocles]

It looks to me as though there is a bit of a discrepancy between the thread title and the question text. At present (leaving aside the long term philosophy of "will we ever be able to ...")

1. Yes, we could definitely build a computer that would contain all of our present scientific knowledge and "work out the questions we ask it" in terms of retrieving and uncritically presenting material from its databank. That would simply amount to a particularly good library index or search engine.

2. Yes, we could definitely build a computer that would "do our experiments" in the sense of being a laboratory assistant that would execute and record the results of any measurements that we required for a particular project, supposing that we were to give it precise instructions. Nearly all modern scientific instruments are under some sort of computer control anyway.

3. No, we could not at present build a computer that would think out and draw up the experimental design for "new" science, although a lot of the work in artificial intelligence is directed at this sort of goal, and we have some idea of how to proceed about building this sort of process into computers in a small way. Whether the grand goal of -- building genuine intelligence or curiosity into a computer is possible or likely to be achieved -- is the philosophical question that I said I would leave aside in my first sentence.

[CliffordK]

Computers can be built to harvest and consolidate data.  You might look at Watson and Jeopardy, so presumably you could make a computer that was very good at Question and Answers.  If you look at some of the "armchair science" on this website, you would notice that each person brings a little different perspective into their discussion, something that might be difficult to capture with your computer.  Also, keep in mind that there are certain predictable types of mistakes that Watson might make, especially with multi-part questions where Watson might overemphasize the wrong part of the question.  Watson, of course, also is designed to come up with one-word answers.  Could it write a whole new paragraph without plagiarism?

As far as running experiments.  Once the experiment is designed, a computer can help a lot.  For example, once a multi-step process for DNA analysis is designed, it could be computerized so that the sample is put in one end and the analysis is read at the other end. 

But, that is different than asking a new question which requires working out an experimental design de-novo.

AI has progressed forward in leaps and bounds, and undoubtedly feats that were never considered possible will be routinely done by computers and AI by the end of this century.

Hmmm....  thinking about answers...  Damocles and I have independently come up with different, but similar answers.  Maybe we aren't so "individual" after all.