Does Computer Science Need Computers?

It sounds like a riddle, or maybe a provocation: does computer science actually need computers? The question, explored in a recent piece from Quanta Magazine, gets at something that surprises many people outside the field. A large and influential part of computer science has almost nothing to do with machines. It is mathematics, done with pencil, paper and proof.

A discipline born before its hardware

The historical answer is the most striking one. The foundational ideas of computer science were worked out before general-purpose electronic computers existed. In the 1930s, logicians and mathematicians were wrestling with a question posed in the abstract: what does it mean for a problem to be solvable by a mechanical procedure? Alan Turing’s answer was an imaginary device — an infinite tape, a read-write head, a finite set of rules — that no one ever intended to build. Alonzo Church approached the same territory through a formal system of function definitions. Neither needed a machine. They needed a definition.

What emerged was a theory of computation that treated “computing” as a mathematical object rather than an engineering artifact. Turing’s machine remains the standard reference model for what is computable at all, and his proof that some problems are undecidable — that no procedure, however clever, can solve them — is a result about logic, not about silicon. Faster chips do not make undecidable problems decidable.

Theory that outruns the hardware

That pattern continues today. Computational complexity theory asks how much time, memory or randomness a problem inherently requires, and its central open question — whether P equals NP — is a mathematical conjecture. It cannot be settled by running an experiment, no matter how large the data center. Cryptography rests on similar reasoning: the security of the systems protecting online banking depends on beliefs about the difficulty of certain mathematical problems, arguments made in proofs rather than benchmarks.

Even quantum computing, one of the most hardware-hungry frontiers in the field, has a thriving theoretical wing that long predated usable quantum machines. Algorithms were designed, and their advantages analyzed, for devices that did not yet exist and in some cases still do not exist at the required scale.

But machines change the questions

None of this means computers are incidental. Real hardware has repeatedly reshaped what theorists care about. The rise of enormous datasets pushed researchers toward streaming and sublinear algorithms, which must produce answers without ever reading all the input. Parallel and distributed machines spawned entire subfields concerned with coordination and failure. The peculiar economics of memory hierarchies inspired cache-aware analysis. And the empirical success of deep learning has confronted theory with a puzzle it did not choose: methods that work far better in practice than existing explanations predict.

There is also a practical dimension. Machines serve as laboratories, letting researchers test conjectures, search for counterexamples and hunt for structures too large to examine by hand. Computer-assisted proof has become a legitimate tool in mathematics itself.

Two disciplines under one name

The honest answer, then, is that “computer science” describes two intertwined enterprises. One studies computation as a natural phenomenon and a mathematical abstraction; it would survive if every machine on Earth were switched off. The other builds, measures and improves real systems, and would be meaningless without them.

The field’s name emphasizes the hardware. Its deepest results often do not — which is precisely why the question is worth asking. Read More