Crazy: Scientists Compute With Human Brain Cells

Quick Overview

The video concludes that while computing with human neuron cultures, as demonstrated by Cortical Labs' CL1 device and the recent Pong experiment, offers incredible energy efficiency (100,000 times less energy than AI systems), significant challenges remain regarding the durability (lifespan of months), upscaling, and reproducibility of these biological systems, making the advent of sentient, artificially grown human brains highly unlikely in the near future.

Key Points: Cortical Labs' CL1 biological computer uses real human neurons grown on a silicon chip, running on 100,000 times less energy than contemporary artificial intelligence systems (0:01-0:11). The system, which costs $35,000, allows researchers remote access to the brain organoids via a Biological Intelligence Operating System (bioOS) (0:30-0:40, 0:55-1:02). Researchers successfully trained a cluster of live neurons to play the game Pong, achieving 80% accuracy by having the neurons update the paddle position based on electrical stimuli encoding the ball's location (1:08-1:23). Three major challenges currently face this technology: Durability (neurons only live for a few months), Upscaling (difficulty in scaling control), and Reproducibility (difficulty in consistently growing and training them) (3:25-3:39). The speaker highlights that these challenges contrast sharply with traditional neuromorphic chips, which emulate learning but lack the flexibility and self-programming inherent in biological neurons (2:35-2:51). Potential future applications include medical uses like aiding patients with brain damage or dementia, and cognitive augmentation, though the ethical considerations are vast (3:01-3:14, 4:16-4:25). The video promotes Ground News, a news aggregation platform that summarizes articles and checks media bias/factuality, offering a 40% discount via a link/QR code (5:20-6:26).

Context: The video presents a science news segment by Sabine Hossenfelder, focusing on the recent advancements in 'Human Neuron Computing' or 'Wetware Computing,' exemplified by Cortical Labs' CL1 device. This technology integrates living human neuron cultures onto microelectrode arrays to perform computations, offering extreme energy efficiency compared to digital AI. The discussion centers on a published study demonstrating these neurons learning to play Pong and outlines the significant technical and ethical hurdles preventing these biological computers from achieving general artificial intelligence.