The first artificial intelligence possesses 'curiosity' just like humans

Recently, scientists have succeeded in 'grafting' curiosity into artificial intelligence (AI), thereby motivating AI to explore itself within a virtual environment. This can be considered a new step in the process of bringing AI into the real world.

An AI does not work

Researchers at the University of California, Berkeley have created an AI that possesses natural curiosity . They proved that their experiment succeeded after the AI played Super Mario and VizDoom.

Scientists have created an AI that possesses natural curiosity.

It is easy to see that when the AI is not equipped for its curiosity, it is constantly bumping into the walls, while the AI of the University of California (DID) is observing, exploring and learning from. surroundings. Pulkit Agrawal, a doctoral student at the DHS, and also a member of the project development team often compared the AI to the children. They always create random 'experiments' and that is how curiosity helps them to explore the world.

Is this a solution to AI problems?

At the present time, most AIs are trained by the 'Reinforcement Learning' method - specifically, AI will be 'rewarded' after completing a designated goal or task. This method is considered to be an effective tactic to train AI to accomplish specific tasks, but it is not really useful when scientists want AI to be proactive and not just follow orders. . Implanting 'curiosity' into successful AI is really a new step, bringing us closer to adopting AI into practice, enabling them to actively help solve global problems.

However, from another perspective this is a real concern. As Elon Musk and Stephen Hawking have argued, AI can become a danger to humankind. And we really need to consider the possible consequences after creating intellectual improvements for AI, supporting its learning process - a process that may not be complete yet. I got it.