'Super Mario' AI Experiment Brings Nintendo's Plumber To Life, Literally

Scientists Literally Bring Mario To Life To Research Thinking Machines
|

While wearable screens and the Internet of Things are getting all the hype in the tech world right now, both rely to some extent on the development of truly clever artificial intelligence.

So - for that matter - do self-driving cars, helpful personal assistants, the future of financial services and virtually every other innovation you can think of.

Which is why it's so important that we teach a computer to play Super Mario Bros.

A study at the University of Tübingen's Cognitive Modelling Group in Germany is doing exactly that in order to expand what we know about how AIs learn and evolve.

The experiment uses a modified version of Mario in which the plumber is effectively self-aware - able to explore his world, learn about it and chat to researchers with Carnegie-Mellon's Sphinx speech recognition toolkit.

That means scientists can 'talk' to Mario, ask him to explore certain things and affect his mood in ways that - in turn - change how he explores his world. Mario won't know that a giant bullet will kill him until the researchers tell him - or he finds out the hard way.

Interestingly this isn't entirely new territory for AI researchers - a competition for computers able to play Mario like humans has been around since 2009 and various other attempts have been made in the meantime.