Introduction
The comparison between biological systems (real neurons) and computer algorithms has been a fascinating topic for scientists. A recent study yielded an interesting discovery by comparing how lab-grown neurons and advanced computer programs learn to play a simple game. The research used a system called DishBrain, which allows neurons in a lab dish to "play" a video game through electrical signals.
Exclusive
The researchers studied two types of brain cells (neurons):
- Human brain cells grown in the laboratory
- Mouse brain cells grown in the laboratory
These cells were compared with three different computer programs using artificial intelligence. The game chosen was a simplified version of the classic Pong, similar to table tennis, where a "racket" must hit a virtual "ball."
Important terms:
- Rally: Sequence of successful hits of the ball without errors
- Aces: When the ball is not hit right at the beginning of the play (similar to tennis)
Results
Brain cells demonstrated surprising performance:
- They managed to keep the ball in play for longer
- They made fewer mistakes on their first attempt to bat
- They learned faster than computer programs
Most impressively, the neurons achieved this superior performance while receiving far less information than the computer programs. While the neurons received signals through only eight contact points, the computer programs received much more detailed information from the game.
Q&A
The results suggest that the brain (even just a few cells in the laboratory) has more efficient ways of learning than modern computers. This efficiency may be related to how neurons:
- They connect to each other
- Modify your connections with experience
- Adapt to new situations
Furthermore, neurons can do all this while consuming much less energy than a computer.
Conclusion
This research shows us that we still have much to learn about brain function to improve our computers and artificial intelligence programs. The study opens up interesting possibilities for the future:
- Development of more efficient computers
- Creation of systems that combine neurons and electronic circuits
- Better understanding of how our brain learns
The discovery that even a few brain cells can outperform complex computer programs at certain tasks shows us how much nature still has to teach us about information processing and learning.
Reference: https://arxiv.org/pdf/2405.16946
