• 0 Posts
  • 178 Comments
Joined 2 years ago
Cake day: October 25th, 2023
  • voracitude@lemmy.worldtoAsklemmy@lemmy.ml•What's the worst you ever hurt yourself as a kid?
    2·
    6 days ago

    I was skating at the local ice rink, and tripped. I think it was an Olympic-sized rink, but that might have been the pool… it was pretty big, anyway. I bounced my head off the ice so hard I saw blue and red fireworks (only time that’s ever happened to me), and slid half the length of the rink on my face wherein I crashed into the barrier. That shit hurt, I still remember it vividly 30 years later, but luckily nothing broken. My mum was simultaneously aghast, and relieved and amazed I wasn’t more injured. She was convinced I’d have fractured something in my face when she saw me fall.

  • voracitude@lemmy.worldtoTechnology@lemmy.ml•A Musician’s Brain Matter Is Still Making Music—Three Years After His Death
    13·
    28 days ago

    This wasn’t “his brain matter”, these were “neuronal organoids” (clumps of neurons) grown from harvesting white blood cells and turning those into stem cells. Then the clumps were networked together with a literal wire to conduct signals between them, for timing.

    Usually in organoids networks the wire delivers either regular, repeating inputs (“clean” pulses) as a reward for succeeding a task, or a random signal (“noise”) for failure; this is how they’re “trained” to play Pong for example:

    In more advanced closed-loop setups, organoid cultures are embedded within simulated environments that allow them to “interact” in a game-like world. By using high-density multielectrode arrays (MEAs) to deliver patterns of electrical signals, researchers can create closed-loop feedback systems that enable organoids to process and respond to certain inputs (Kagan et al. [2022]). For instance, in one experiment, monolayer neuronal cultures were given sparse sensory feedback about the consequences of their actions within a simulated game. The organoids displayed short-term memory by organizing themselves in goal-directed ways, effectively learning to complete simple behavioural tasks. This capability, made possible by reinforcement learning, allows organoids to adapt based on feedback, akin to how a human brain might learn from trial and error.

    (https://www.cell.com/neuron/fulltext/S0896-6273(22)00806-6)

    These same methods are being used to train organoids as Machine Learning compute substrates, because they’re much more efficient than silicon: https://aapsopen.springeropen.com/articles/10.1186/s41120-025-00109-3