15:29Machine-translated from Azerbaijani by AIRead in Azerbaijani
Google Fully Maps a Fruit Fly’s Nervous System for the First Time
Google Research scientists have created the first complete map of the brain and central nervous system of an adult male fruit fly. Announced on September 3, the project was developed with the participation of the HHMI Janelia Research Campus and other researchers. Fruit flies are widely used in scientific experiments. The new map covers both the brain and the rest of the central nervous system.
As part of the project, artificial intelligence combined millions of 2D images to create a 3D model of the neurons. This allowed researchers to reconstruct the structure of more than 166,000 neurons and the connections between them. The model, called MaleCNS v1.0, is a connectome. In other words, it is a digital map showing how nerve cells are connected to one another. According to Google, the project could accelerate research into how the brain works. Software engineer Alex Wormuth began training the model in the game Doom on September 6. Each frame in the game sends a signal to sensory neurons. The neural activity produced in the model is then converted into commands that control the game. When the character takes damage, a separate signal is sent to 2 PPL101 dopamine cells to reinforce the learning process. The project’s code has been released publicly, and the training of 166,700 neurons can be watched live.
Jessica Paquette also tested the same neural map in Super Mario 64. In a short video published on September 7, Mario repeatedly moves forward and crashes into a wall. The code for this project is also open source. The creator said that the code was made 100% with GPT Astra and solely for fun. So far, the published videos do not show that the model has successfully learned to play the games. Fruit flies have been used in genetics and neuroscience research for more than 100 years. In the new experiments, the game is not being played by a real insect. The game’s commands are linked to activity generated within the fruit fly’s digital neural map. Therefore, the Doom and Super Mario 64 tests are primarily examining whether the model, which has more than 166,000 neurons, can learn behavior within a game.
As part of the project, artificial intelligence combined millions of 2D images to create a 3D model of the neurons. This allowed researchers to reconstruct the structure of more than 166,000 neurons and the connections between them. The model, called MaleCNS v1.0, is a connectome. In other words, it is a digital map showing how nerve cells are connected to one another. According to Google, the project could accelerate research into how the brain works. Software engineer Alex Wormuth began training the model in the game Doom on September 6. Each frame in the game sends a signal to sensory neurons. The neural activity produced in the model is then converted into commands that control the game. When the character takes damage, a separate signal is sent to 2 PPL101 dopamine cells to reinforce the learning process. The project’s code has been released publicly, and the training of 166,700 neurons can be watched live. Jessica Paquette also tested the same neural map in Super Mario 64. In a short video published on September 7, Mario repeatedly moves forward and crashes into a wall. The code for this project is also open source. The creator said that the code was made 100% with GPT Astra and solely for fun. So far, the published videos do not show that the model has successfully learned to play the games. Fruit flies have been used in genetics and neuroscience research for more than 100 years. In the new experiments, the game is not being played by a real insect. The game’s commands are linked to activity generated within the fruit fly’s digital neural map. Therefore, the Doom and Super Mario 64 tests are primarily examining whether the model, which has more than 166,000 neurons, can learn behavior within a game.