May 16, 2021

Artificial intelligence, dragons and dungeons | Trends

Artificial intelligence, dragons and dungeons | Trends


You do not have to be very smart to beat Kasparov to chess. We do not say it. The insinuates Sally Davies This social anthropologist, who studies the way people understand artificial intelligence and our efforts to achieve a system capable of emulating our abilities, has not been impressed by the great victories of the algorithms.

The victory of Deep Blue against Kasparov? The Watson one in Jeopardy? The AlphaGo about Lee Sedol? The one of Libratus in front of the world elite of the Poker? Paparruchas. From Davies' point of view, the games we choose to prove that artificial intelligence is approaching us intellectually are the same: "They tend to be very limited, with fixed objectives and well-defined paths until victory and defeat." Not like Dragons and Dungeons.

The stories that are created on the RPG boards have that I do not know what the other games do not have. The narrative is constructed collaboratively, with an open end and from different points of view. "First, the intelligence has to work in different scenarios: Dragons and Dungeons players can inhabit a lot of characters in the games and the individual player can change roles: the fighter, the thief, the healer … Those who investigate in artificial intelligence know that it is super difficult to get a well-trained algorithm to apply their knowledge in different scenarios, even when the variation is minimal, "explains Davies.

So, why do we insist that our algorithms win corseted games? The anthropologist quotes a former Berkeley professor, Robert Wilensky. The teacher, who died in 2013, devoted his entire career to the research of artificial intelligence. His theory is that this mania is simply a professional deformation. "All of them were essentially formative mathematicians, and mathematicians do two things: prove theorems and play chess, so they said, 'If you can prove a theorem or play chess, you must be smart'."

However, the limitations of the usual winners do not end in the poor adaptability to variable scenarios. "Dragons and Dungeons reminds us that intelligence has a body," says Davies. Players need to act so that their avatars do so and react to the development of the story, not only on the strategic level, they also respond emotionally to the course of events: fear, stress, frustration … "Recent research suggests that physical interactions They are crucial in the way we understand the most complex mental concepts, but we pay minimal attention to the corporality of artificial agents and how this could affect the way they learn and process information, "laments the anthropologist.

The last gift of the popular RPG to the advance of artificial intelligence is its social side. Davies admits the validity of learning models that reinforce successful strategies with rewards, but considers them insufficient. His alternative recipe has something of the idea of ​​not giving the cane, but teaching to fish: the teacher must accompany the student in his learning of a new skill. "In the case of games, this cooperation is channeled through history […]. Instead of beating opponents in games, we could learn more about intelligence if we tried to teach artificial agents to play together as we do: as paladins and elves. "

.



Source link