Home Sciences Spanish scientists teach bacteria to play Tres en Raya

Spanish scientists teach bacteria to play Tres en Raya


Researchers from the Higher Council for Scientific Research (CSIC) have managed to enable bacteria to play Tic Tac Toe and reach the level of experts. They have genetically engineered microorganisms so that they are capable of learning algorithms on their own and competing with human players, something never seen before.

An international investigation led by the Spanish scientist Alfonso Jaramilloof the De novo Synthetic Biology Laboratory, I2SysBiobelonging to the CSIC and the University of Valenciahas developed a methodology that allows bacteria, for the first time, to play Three in a rowan entertainment that consists of being able to place three tiles aligned on a board.

In this pioneering technological development, the game does not take place on a board, but on a Petri disha round glass container used in scientific research to observe different types of samples, both biological and chemical, enclosed within the plate.

The learning matches take place between humans and bacteria and, unlike the classic game, they can last for days, because it is a much more complex experience: never before has the training of bacteria gone this far.

The experience emulates another technological innovation achieved in 1952 by the first computer in history, Eniac (Electronic Numerical Integrator And Computer), built by the University of Pennsylvania, which took up an entire room and could solve 5,000 additions or 300 multiplications in 1 second.

AI demo

Eniac managed to play OXO, which could be the first electronic version of Tic Tac Toe, a computer feat that represented one of the first experimental demonstrations of what the Artificial intelligencethe technology that aims to replicate human cognitive abilities in a machine.

Eniac was programmed to play and worked through vacuum or thermionic valves, used to amplify, switch, or modify an electrical signal by controlling the movement of electrons in an empty space at very low pressure.

What the CSIC investigation has achieved, in which it has also participated Adrian Racovitafrom the same CSIC-University of Valencia Institute, together with researchers from the universities of Warwick and Keele (both in the United Kingdom), is to apply synthetic biology so that these microorganisms are capable of learning algorithms by themselves, of playing Tic Tac Toe and even never losing to a human player.

Read:  For a man born blind, numbers are cubes and letters are two-dimensional squares.

As the authors of this research explain in their article, the strains trained in the framework of this research can learn to master board games such as Tic Tac Toe, even starting from a state of total ignorance.

Autonomous Learning

The highlight of this research, as Jaramillo explains to Tendencias21, is that it provides a general genetic mechanism for autonomous learning of decisions in changing environments.

The team has developed a adapted genetic circuit so that the bacteria remember the behavior they adopt during the game, recognize if it has been a useful move and can change their attitude in the following moves, thus improving their competence for the game.

The secret of this development, adds Jaramillo, are chemical compounds that allow bacteria to recognize the different plays that occur on the board through the cell fluorescencewhich indicate the success or error of your moves.

At the beginning of the game, the bacteria always lose, as they are completely unaware of what they are up against, but as the learning games go on, they learn the mechanics of the game, they recognize the successes of the human player, they learn from their movements and they develop, after an arduous learning process, the ability to compete with the complex human intelligence.

expert level

Learning each movement, which is carried out by each bacterium itself, takes him at least 15 hours, which means that he loses seven games in a row until he reaches the expert level to the eighth challenge.

The bacteria, starting each game after a first move by the human player, always start at a disadvantage, and the maximum level of intelligence is reached when they are able to never lose a game.

Read:  First runaway black hole detected in the Milky Way

Comparing this development with the feat of Eniac of the last century, who was not able to learn by himself, as our bacteria do, we can say that we have advanced further, because we have not had to program the bacteria: they did it automatically from your experienceJaramillo explains.

And it specifies: Although current computers are also capable of learning to play, our research shows that the power achieved with computers to learn to play chess, the Chinese strategy game Go, or Tic Tac Toe, can also be achieved with genetic circuits able to autonomously learn decision-making in complex environments.

This technological development is not entertainment, but promotes the development of Artificial Intelligence in living cellssomething that until now did not exist, indicates the Spanish researcher.

never seen

He adds that the calculations carried out by bacteria to learn to play Tic Tac Toe is something that never seenwhich will encourage new research in this promising scientific field.

Jaramillo also points out that this development allows us to think about the possibility that in the future the concepts proposed for bacterial engineering can be generalized, so that they communicate with electrical impulses, instead of through chemical compounds and fluorescence, which It would considerably reduce the response and learning time.

One could even think of the possibility of using these electrical networks of bacteria to build mini brains with cognitive abilities similar to those of humans, concludes Jaramillo.

The results of this research have been published in bioRxiv, the largest non-profit, open access online repository, collecting preprints of life science-related articles that have not yet been peer-reviewed. The researchers are in the process of publishing their work in some of the most important scientific journals.


Engineered gene circuits with reinforcement learning allow bacteria to master gameplaying. Adrian Racovita et al. bioRxiv, April 25, 2022. https://www.biorxiv.org/content/10.1101/2022.04.22.489191v1

Previous articleThey build a seawater desalination plant the size of a suitcase
Next articleThe fall in pollution in Europe since 1980 favors more violent hurricanes