Machine Learning and Quantum Machine Learning: Teaching Computers to Find Patterns
Imagine that you want to teach a computer how to tell the difference between a picture of a cat and a picture of a dog.
One way would be to give the computer a long list of rules:
“Dogs may have floppy ears. Dogs often have long noses. Cats may have pointed ears. Cats may have whiskers.”
But this would be difficult because animals look very different. Some dogs have short noses, some cats have floppy ears, and pictures can be blurry or taken from unusual angles.
Instead, we could show the computer thousands of pictures. We would label each picture cat or dog. The computer would study the examples and look for patterns that help it make a prediction about a new picture.
This is the basic idea behind machine learning.
Machine learning is a type of artificial intelligence, or AI, that allows computers to learn patterns from data. Data can include numbers, pictures, words, sounds, videos, or information collected by sensors such as cameras or radar.
A machine-learning system usually works in three main steps:
1. Collect examples. The computer receives data.
2. Find patterns. The computer uses mathematics to discover relationships in the data.
3. Make a prediction. The computer uses what it learned to make a guess about new information.
The computer does not “understand” the world exactly as people do. It uses mathematics to recognize patterns. Its predictions can be useful, but they can also be wrong.
Example 1: A Music Recommendation
Suppose you listen to several songs and give each one a rating from one to five stars.
A music app might notice that you often enjoy songs with:
• Fast rhythms
• Strong bass
• Certain singers
• Similar musical styles
The app can use these patterns to recommend a new song. It might say:
“Because you enjoyed these songs, you may also like this one.”
The computer is not reading your mind. It is comparing information about your choices with patterns found in the data.
However, the recommendation may not always be correct. You might actually like Jazz but have been reviewing Classical or Punk Rock for the last few weeks. Today, maybe you want to hear something completely different (like Jazz). Machine learning makes predictions based on patterns—it does not know exactly what you will choose.
What Is Quantum Computing?
Most computers use tiny pieces of information called bits. A bit can have one of two values:
0 or 1
You can imagine a bit as a light switch that is either off or on.
Quantum computers use quantum bits, or qubits. Qubits follow the unusual rules of quantum physics, the science that studies extremely small things such as atoms and particles.
A qubit can exist in a special quantum state that is different from an ordinary bit. Qubits can also become connected in a quantum relationship called entanglement.
These properties may allow quantum computers to solve certain specialized problems in new ways. However, quantum computers are not simply “super-fast versions” of ordinary computers. They are designed for particular kinds of difficult problems.
Quantum computers are also still developing. Qubits are fragile and can be affected by noise and errors. For many everyday jobs—writing a report, watching a video, playing a game, or sending an email—a regular computer is still the better choice.
What Is Quantum Machine Learning?
Now, we're beginning to bring it home. Quantum machine learning, often called QML, combines ideas from machine learning and quantum computing.
Machine learning asks: “How can a computer learn useful patterns from data?”
Quantum computing asks: “How can we use the unusual rules of quantum physics to solve certain problems?”
Quantum machine learning asks: “Can quantum computers help machine-learning systems solve some problems in better or more efficient ways?”
Scientists are still researching this question. Quantum machine learning is exciting, but it is still an emerging field. Researchers have not yet shown that quantum machine learning is better than ordinary machine learning for most practical problems. This said, the promising is ... considerable.
Example 2: Searching for New Medicines
Scientists may study huge numbers of molecules while searching for new medicines. Each molecule can have many properties, and the possible combinations can become extremely complicated.
Machine learning can help scientists examine data and predict which molecules may be promising, as was done in the search for responses to COVID-19.
In the future, quantum machine-learning methods might help researchers study certain complicated patterns more efficiently. For example, a system might help answer:
“Which molecules are most likely to interact with a particular protein?”
The computer would not create a medicine by itself. Scientists would still need to perform laboratory experiments and carefully test whether a possible medicine is safe and effective.
The Big Idea
Machine learning helps computers learn patterns from examples and use those patterns to make predictions.
Quantum machine learning explores whether the special abilities of quantum computers can help with some machine-learning problems.
A simple way to remember the difference is:
Machine Learning: Learning patterns from data.
Quantum Computing: Using quantum physics to perform certain kinds of calculations.
Quantum Machine Learning: Exploring how quantum computing might improve some machine-learning tasks.
Machine learning is already used in many tools today. Quantum machine learning is still being developed, but it may eventually help scientists study difficult problems in medicine, chemistry, energy, climate science, and other fields.
The most important lesson is this:
Computers can learn patterns from data, but humans must decide how to use those predictions responsibly.
BSF takes the promise of quantum computing and quantum machine learning seriously. Consistent with our emphasis on math literacy, we will offer programs on these topics. Interested families can be put on a mailing list and be kept abreast (here).