Algorithms suffuse our daily lives. They allow us to search Google with relative success, fly us across the globe, and helpfully recommend books we may like to read. But just like the humans that program them, algorithms are fallible.

The potential implications of this fallibility were starkly underlined on Wednesday when a glitch in one of the trading algorithms deployed by US brokerage Knight Capital led to a malfunction that has ultimately cost the company $440m and left it scrambling for a deal to bolster its capital base. The precise cause of the problem remains a matter of speculation.