Studyledge

Artificial Intelligence: Question 6

Syllabus 18.1

Multiple choice A2 1 mark

A speech-recognition company builds a system that converts spoken audio into written text. The system uses an artificial neural network with 40 hidden layers, trained on millions of labelled audio clips paired with their correct text transcriptions. Each successive hidden layer learns to recognise increasingly complex patterns in the audio, building on the simpler patterns learned by the layer before it.

Which term most precisely describes this system, given that it uses an artificial neural network with many hidden layers to learn increasingly complex patterns from the data?

Choose an answer to check it, then compare with the worked solution below.
Show worked solution Hide worked solution

Worked solution

Identifying deep learning

The key detail in this scenario is that the artificial neural network is built with 40 hidden layers, and that each successive layer learns increasingly complex patterns, building on the simpler patterns learned by the layer before it. This. A neural network with many hidden layers, used to learn increasingly complex, layered patterns in the data, is exactly what deep learning refers to.

Note that “deep learning” describes the network’s structure (many hidden layers), not the type of training data it uses. This particular system also happens to be trained on labelled data (so it is also an example of supervised learning). The two are not mutually exclusive.

Why the other options are wrong

  • A (unsupervised learning): unsupervised learning trains on data with no known correct output attached. Here, every audio clip is explicitly labelled with its correct transcription, so this is not the case.
  • B (reinforcement learning): reinforcement learning involves an agent taking actions and learning from reward or penalty feedback received through trial and error. There is no agent taking actions and no reward signal described here, only a fixed set of labelled training clips.
  • D (a regression method): regression methods predict a continuous numeric value from input data (for example, a price or a temperature). Converting audio into written text is not a numeric-value prediction, so this is not a regression task.

Final answer

C. This is deep learning, because the system uses an artificial neural network with many hidden layers to learn increasingly complex patterns in the data.