The ultimate goal of AI is to transition from supervised learning to thinking like humans.

When the robot goes to sleep, what will it dream of? Obviously, the Atari game.

People's nighttime sleep (or daytime snoring) helps to consolidate memory and transform short-term memory into long-term memory. By stabilizing, enhancing, and integrating three different processes, your brain can turn memory into a more organized file system so you can recall it more easily in the future.

First, the stabilization process helps humans encode a piece of memory in 6 milliseconds. Then, the brain will enhance memory in minutes, hours, and even the whole day, and consolidate it into long-term memory. Finally, the integration process, the brain will add new memory fragments to the existing memory, this process is a bit like a file system, rather than a file cabinet.

The ultimate goal of AI is to transition from supervised learning to thinking like humans.

Researchers hope that robots will eventually be like humans. DeepMind, a subsidiary of Google (microblogging), has achieved great success in classic video games. Games such as "Breakout" and "Asteroids" not only teach artificial intelligence (AI) not simply to sort the game, but also lay the foundation for today's supervised learning technology. In humans, you must learn to climb before you learn to walk. On robots, you must first defeat the game before conquering face detection or cancer research.

Although DeepMind's technology is advancing, it still can't beat humans in more complex games, such as StarCraft or Labyrinth. When we humans dream of embarrassing situations or threats, AI dreams of rearranging all the chapters of these games in order to pave the way for victory, and they repeat the process until they become "experts."

The goal is to let AI learn through experiments like humans. From supervised learning (AI analysis of data and finding patterns) to unsupervised learning, including teaching robot experiments and analyzing the effects of different behavioral processes on outcomes. This kind of learning is more time consuming because there are countless variables. This provides an ideal solution for AI when it is inactive or dreaming.

This is still an emerging field of research in the AI ​​field. But so far, researchers have reported impressive speeds, and the speed of supervised learning has increased tenfold. Although it is still experimental, preliminary research shows that robots focus on the work scene, at least in dreams, which may be beneficial for future AI evolution.

Subway Escalator

Subway Lift,Subway Elevators,Subway Escalator,Subway Lift Elevators

XI'AN TYPICAL ELEVATOR CO., LTD , https://www.chinaxiantypical.com

Posted on