Jeffrey Shainline: Neuromorphic Computing and Optoelectronic Intelligence | Lex Fridman Podcast #225

Lex Fridman and Jeffrey Shainline take a deep technical dive into computing hardware, starting from semiconductor physics, transistors, and Moore's law before turning to superconductivity and Josephson junctions. Shainline explains his core thesis: electrons excel at computation while light excels at communication, and why integrating compound-semiconductor light sources with superconducting electronics at 4 Kelvin could enable brain-scale neuromorphic systems. They explore first principles of the brain, including its fractal, power-law structure across space and time, memory, plasticity, and modular architecture. The conversation closes with Shainline's speculative paper extending Lee Smolin's cosmological natural selection, arguing that intelligence and technology may be what our universe's fine-tuned parameters were selected to produce.

• Shainline argues silicon's dominance is enabled by physics itself, it 'almost as if it was there waiting for us to discover it' rather than being purely invented.
  00:10:19
• He flatly predicts superconductors will NOT replace semiconductors for digital computing because all-things-considered semiconductors work better at the system level.
  00:34:44
• He says it wouldn't take much for superconductors to beat silicon, the problem is they simply aren't better at building a whole digital computer.
  00:36:48
• Shainline positions himself at the radical end of neuromorphic computing: abandoning transistors entirely and designing hardware from first principles of the brain.
  00:59:37
• Contrarian claim: he says neuroscience is far more mathematically rigorous and 'theoretically complete' than the common refrain that 'we don't understand the brain.'
  01:24:21
• He reveals superconducting single-photon detectors cut required light levels by three orders of magnitude, so 'you can have the world's worst light sources.'
  01:37:16
• He admits his life goal is to 'design things that I hope will think about themselves', systems with self-reflective consciousness.
  02:30:15
• Speculative bombshell: a technological civilization could turn a single asteroid into a trillion black holes (offspring universes), implying WE may have come from technology, not a star.
  02:33:51

• State-of-the-art transistor gates are around seven nanometers, just a few tens of atoms along the conduction pathway.
  00:07:14
• A 300mm silicon wafer is 'as big as a pizza' and can hold around a thousand processors before being diced up.
  00:09:48
• Flux-on current pulses from Josephson junctions can propagate at roughly a third of the speed of light and switch in tens of picoseconds.
  00:29:31
• Superconducting devices can hit hundreds of gigahertz versus ~1-3 GHz for conventional processors.
  00:30:33
• Each neuron in your brain makes roughly 10,000 connections, which is impractical to drive with electrons over wires.
  00:50:51
• Brain connectivity follows a power law in space (like gravity or electromagnetism), giving it fractal, scale-invariant structure.
  01:03:43
• The cortex (~10 billion neurons) unrolled is about the size of a pizza; the hippocampus has ~100 million densely, almost randomly connected neurons.
  01:09:54
• Quantum computing operates near 50-100 millikelvin, far colder than the 4 Kelvin Shainline's superconducting systems need.
  01:38:18
• The cosmic microwave background sits at about 2.7 Kelvin, colder than the 4K his hardware requires, so deep space is naturally suited to it.
  01:38:49
• Compressing about 10 kilograms into a tiny volume could form a black hole likely to inflate into its own offspring universe.
  02:35:25