Scientists train brain cells in hybrid chips to play ping-pong at a much faster learning rate than AI

Whether it's Terminator or Halo, science fiction works seem happy to cast AI as an exceptionally intelligent and dangerous presence. But you may not imagine that the potential of the human brain may be more powerful than AI.

According to New Scientist, a team in Australia has been trying to grow brain cells. Now, some of the human brain cells in the dish perform much better than AI in video games, and the learning time is very short.

Cortical Labs is a company dedicated to integrating biological neurons with traditional silicon-based computing hardware. They culture brain cells on microelectronic arrays to stimulate the cells. These hybrid chips are said to be able to learn and reorganize themselves to solve problems, such as learning simple electronic table tennis games.

According to Cortical Labs, IT usually takes 90 minutes to learn this ping-pong game, and this "DishBrain" (the name of the hybrid chip) can manage to complete in 5 minutes. But the researchers note that once both are properly trained, a good AI will still beat the brain cells.

The team did this by using electrical signals to communicate the ball's position during the game to "DishBrain," which could then move the plate to collide with the ball. Although it may seem like a very crazy science experiment, it is still much worse than humans, at least I am sure that the calculation of this game is faster, and does not require the help of a team of scientists.

Brett Kagan, chief scientific officer at Cortical Labs, told New Scientist that "we think it's fair to call them machine brains," going on to explain that when brain cells move the plates in a game, they think they are the plates themselves. "We often call them living in the Matrix." He explains, "When they play the game, they believe they are boards. ”

Now AI has become a development trend in various industries. AI can turn text into realistic images, or encode things, and Bethesda wants to use better AI in games. But we're also excited to see how terrible hybrid chips like DishBrains can do in development.