Human Brain Cells Powering a Computer Learn to Play 'Doom'

Australia - Ekhbary News Agency

Human Brain Cells Powering a Computer Learn to Play 'Doom'

In a groundbreaking development, scientists at Australia-based Cortical Labs have successfully trained a biocomputer, comprised of lab-grown human brain cells, to master the notoriously challenging video game 'Doom'. This remarkable feat signifies a major leap forward in the field of biocomputing, demonstrating the potential for organic systems to perform complex, goal-oriented tasks in real-time. The breakthrough is seen as a pivotal moment in the quest to develop a new era of hybrid organic technologies that merge biological intelligence with silicon-based systems.

"This was a major milestone, because it demonstrated adaptive, real-time goal directed learning," stated Brett Kagan, Cortical Labs Chief Scientific and Chief Operations Officer, in a recent video announcement. This achievement transcends the simple novelty of organic matter interacting with digital media; it represents a profound insight into the adaptability and learning capabilities of neural networks, whether biological or artificial.

The journey to enabling brain cells to play 'Doom' has been a lengthy one. The company's earlier work, highlighted in 2021, involved a biocomputer named 'DishBrain'. This pioneering system utilized approximately 800,000 human nerve cells, intricately connected to a small processing chip. This chip was designed to interpret and direct electrical activity, mimicking the foundational principles of conventional silicon-based computing. To illustrate the potential of 'DishBrain', engineers initially trained it to play 'Pong', a simpler, 2D arcade game often used as a benchmark for computational neuroscientists due to its requirement for real-time navigation of a dynamic information landscape.

Achieving the 'Pong' benchmark took Cortical Labs over 18 months using their original hardware and software. However, the ambition to tackle more complex challenges led to the development of 'CL1', the system now capable of engaging with 'Doom'. The company proudly bills 'CL1' as the "world’s first code deployable biological computer," underscoring its readiness for advanced applications beyond basic demonstrations.

The transition from 'Pong' to 'Doom' highlights the increasing complexity and utility of biocomputers. While moving a digital paddle was an important proof of concept, the true value lies in tasks demanding more sophisticated cognitive processing. 'Doom' has long served as a ubiquitous test for technological prowess in the computing world, with enthusiasts and major companies alike finding ways to run it on everything from calculators to tractors. The question for Cortical Labs wasn't 'if' they would attempt to run 'Doom' on neuronal chips, but 'when'.

A significant hurdle for 'CL1' was the necessity to process visual input, essentially 'seeing' the game as a human player would. Lacking inherent optical sensors, the engineering team faced the challenge of converting visual data into electrical stimulation patterns that the cultured neurons could interpret and respond to. Remarkably, this complex problem was solved in approximately one week by Sean Cole, an independent developer with limited prior experience in biological computing. The key innovation was the development of a new interface for 'CL1', enabling programming through the widely accessible Python language.

While 'CL1' is not yet a 'Doom' tournament champion – it performs better than random firing but still loses frequently – its progress is noteworthy. Cortical Labs reports that 'CL1' achieved its current level of performance significantly faster than traditional silicon-based machine learning systems. The company anticipates further improvements as its algorithms are refined. The implications of this technology extend far beyond gaming; future iterations of biocomputers could potentially power advanced robotic limbs, manage intricate digital processes, or even contribute to new forms of medical research and treatment. Successfully navigating the 'Doom' benchmark is a strong indicator of the technology's promising trajectory.

Ekhbary News Agency