When Google announced earlier this month that it was launching an artificial intelligence-based search function, it promised that “Google will do your Internet searches for you.” A new feature called AI Overview provides brief AI-generated summaries that highlight key information. A link will appear at the top of the search results.
Unfortunately, AI systems are inherently unreliable. And within days of the AI overview being released in the US, users were learning how to add glue to pizza, eat at least one small rock a day, and learn whether former US President Andrew Johnson earned a college degree between 1947 and 1947. I quickly shared an example of a feature that suggests this. 2012, despite having died in 1875.
Yesterday, Google Search chief Liz Reid announced that the company was making technical improvements to its system.
So why does AI Overview return unreliable and potentially dangerous information in the first place? And what can you do to solve the problem? Read the full story.
—Rhiannon Williams
AI-controlled drones could help find lost hikers faster
When hikers become lost in the rugged Scottish Highlands, rescue teams sometimes send in drones to look for clues about the individual’s route. However, due to the large terrain and limited battery life, it is important to choose the right area to search.
Traditionally, professional drone pilots use a combination of intuition and statistical ‘search theory’ – a strategy with roots in hunting German submarines during World War II – to prioritize certain search locations over others.
Now researchers want to see if machine learning systems can perform better. Read the full story.
—James O’Donnell
