Can living brain cells unlock the next big breakthrough in AI?
Neurosurgeon turned founder Alex Ksendzovsky thinks so.
His startup @BioComputingCo trains AI models that he claims are as much as 27x more efficient by taking advantage of the cheapest, best image processor available: the neuron.
“I have about 500 million years of evidence behind me,” he says. “So our moat is evolution.”
TBC has raised more than $25M and is already in the market with its models. But its CEO is chasing a future straight out of sci-fi: real-time biological computing happening inside data centers.
“We’re bringing something completely crazy to the world, and we’re showing that it works."
On The Upstarts Podcast, we discuss why algorithms derived by brain cells beat ones only inspired by the brain; why models were the best place to start; and the ethics of using living neurons for training AI.
Plus, Ksendzovksy shares his Upstart Moment: giving up a promising career in academia to pursue a high-risk startup.
CHAPTERS 00:00 Introduction 01:12 Building software from wetware 05:09 A college idea to predict the stock market 08:40 The first test from an Airbnb 12:46 Rat neurons and electrodes 17:30 Pitching VCs with no data 24:52 Promising early AI results 31:09 The ethics of using neurons 34:40 ‘Our moat is evolution’ 35:38 A real-time biological computing future
This season of the show is presented by @Rippling 🫡