New Ai writes the math problems that drink millions of steps

The researchers have made a wise intelligence not a form of actual opposition to a period of time. But it is not limited to just a few expenses expected – it thinks that millions of the steps of the future.

Group Guilt with mathematics Sergei Gikov from California Institute of Teclogy (Caltech) made a new commercial condition to solve mathematical problems. If indeed a long-term list; We are speaking one million or more.

Mainly, AI has been able to progress a difficult problem called Andrews-Curtiswhich has analyzed several decades. The suggestion that asks: Do some math photos are always solved using methods allowed, as renewed or destroying stairs?

To make this possible, a new Caltech app was required to find “to find a long-term of the most important and difficulty,” is Sheeper, the maths, the maths on the University and algech statement. “It’s like trying to find your way through the size of the world. These are the longest ways you need to investigate, and only one way to work.”

In the Revieft’s Review Arxiv The Last August is a Data Tuesday, Shehper and employees as they use their own genes that are made to stop the problems with Andrates, which includes algebra. To make a good look, they did not resolve itself. Although this may be seen as an ironicactic, the researchers have imitated what can be done to match what can happen. Despite the opposition of meetings does not make the first one as true, we hit.

Sheffer said: “To introduce some of these aspects, it assures us that we do not pay attention to the original allowance and also helps to make our minds.” Gokov multiplied the mathematical problems in CUBE.

“Can you take a difficult cube, the most difficulty of Rubik and restore his original country? You should try this in consecutive, and you don’t know if you are on the right way to the end.

So how do the Ai doing? Actually, I think outside the box. Following a recommended, research method prompts AI finally eating maths after Math’s math problems after followers. Shehper. “We are recommending the program to do the same thing and keep your knowledge. It has new ways to be better than people can do. It’s learning.”

Algorithm basically the algorithym learned a lot to make a long-term movement, which the researchers were transferred “Super transferred.” In contrast with that, the rotation of the chatppt produces very well.

“When you ask ChatCpt to write a letter, it comes with something. It is not unlikely to be existing with anything special and original. And a good parrot,” says Gurror, “says Gurror. “Our program is better to come with sellers.”

I can think about one first event that would be easier for Ai to be predicted: financially. But when the programs of the current machine has learned something did not meet these exchange for this, researchers describe that their methods may be the same.

Giday observes: “Our points know how to learn.” I thought the group was very important. “Increases, Gokov and friends who had begun to meet a lot of money to hold, makes Their work should have a few computers.

While that the help of this will not appear in our daily life, their work is connected to other algorithms of the algorithms to solve the problems of the people (not to waste development).