Unwise to give AI unfettered access to networks

Given the potential benefits of artificial intelligence (AI), the recent announcement that up to $150 million is to be invested in AI is good news ($150m boost for AI, in bid to solve 'grand challenges'; May 4).

However, AI presents risks as well as benefits for a small country like Singapore.

An open letter by the Future Of Life Institute, titled "Research priorities for robust and beneficial artificial intelligence", noted that "it may be useful to create containers for AI systems that could have undesirable behaviours and consequences in less controlled environments".

AI systems that are able to reprogram themselves present the highest risk, as their behaviour is inherently unpredictable.

For such advanced systems, it would be prudent to restrict access to the Internet and other computer networks.

A set of protocols could help define levels of risk and spell out the corresponding steps to maximise safety.

Otherwise, the tendency would be for researchers to take bigger gambles in the race for better results.

Ben Gibran

A version of this article appeared in the print edition of The Straits Times on May 08, 2017, with the headline 'Unwise to give AI unfettered access to networks'. Print Edition | Subscribe