Professor Sophie Scott starts by explaining that speech recognition is known in science as a "ridiculously difficult question" because speech is incredibly complex.
She emphasises this by playing a sentence in Estonian, and asking the audience to count the number of words they can hear.
She then records a spectrogram to show that what a computer gets when we talk to it is essentially a continuous stream of audio.
A computer then needs to break this up into smaller chunks of information.
This is because it is when you hear the gaps in words that you understand what is being said.
As we all talk differently, a computer has to use an idealised version of speech, and therefore, Sophie explains that what a computer is looking for is a best guess.
Two volunteers are then invited onto the stage to have a go at decoding a stream of speech like a computer.
This clip is from the The Royal Institution Christmas Lecture - The Language of Life.
Teacher Notes
This film could be used to start a discussion on speech recognition; how it is currently used and ways it could be developed and used in the future.
You could ask students to make a list of the devices they can think of that use speech recognition software.
They could then come up with potential ways it could be developed and used in the future.
You could pause the film at the end and ask students to have a go at decoding the stream of speech.
Curriculum Notes
This film is relevant for teaching Physics and Biology at Key Stage 3 in England, Wales and Northern Ireland and at 3rd Level in Scotland, and Cambridge IGCSE Physics.