Foundations

What This Field Actually Is

7 min

Two of the most over-hyped words in technology, stuck together. You would be right to be sceptical.

There is real work here, and there are methods worth learning. There is also a long trail of "breakthroughs" that quietly turned out to be nothing. This track teaches both.

Four different things wear this name

Quantum computers speeding up ordinary AI. The most claimed. The least demonstrated.

Quantum circuits used as the AI model itself. What most current research actually does, and what the next two lessons are about.

Ordinary AI helping quantum computers. Neural networks that spot errors in quantum chips. Unglamorous. Works today. Nobody writes headlines about it.

Quantum models learning from quantum data. Learning directly from a physics experiment or a chemistry simulation, without ever converting to numbers on a disk. The strongest results in the field. Almost never mentioned.

The problem nobody mentions first

A quantum computer cannot open a spreadsheet.

Every number you have has to be converted into the state of some qubits before anything begins, and that conversion takes time.

If loading a million rows takes a million steps, then a "lightning-fast" quantum algorithm that runs on those rows is not lightning-fast at all. You have just moved the cost.

A very large share of quantum AI speed-up claims fall apart at exactly this point. It is the first thing to check, always.

Getting data in is the bottleneck
Basis encodingn qubits for n bits

one bit per qubit

trivial to prepare, no compression

Angle encodingn qubits for n features

one feature per rotation angle

shallow circuits; what most QML papers use

Amplitude encodinglog₂(n) qubits for n features

features become amplitudes

exponentially compact — but preparing it usually costs what you saved

A quantum model cannot read a CSV. Every QML proposal has to answer this question first, and the answer usually determines whether any claimed speedup survives.

What is genuinely solid

Kernels

A real method with proven advantages — on problems designed for it

Expressiveness

Quantum circuits can represent patterns similar-sized ordinary models cannot

Not yet

No advantage shown on a real dataset anyone outside the field cared about

That last line is not an insult. It is simply where things stand, and knowing it is what lets you read a paper usefully instead of anxiously.

Worth remembering

  • Four different things share this name, and conflating them causes most of the confusion.
  • Most current work uses a quantum circuit as the model itself.
  • The strongest results are on quantum data — the corner nobody publicises.
  • Always check the cost of loading the data. Many claims die there.
  • Nothing has yet beaten ordinary methods on a real-world dataset.