This from the same guys who got neurons in a petri dish to successfully play pong. The CL1 is a box with biological neurons inside that can be controlled and programmed with an API. They’ve been working on this for quite a while and just made their first public release.
Its expected the CL1 will be extremely helpful for researching Alzheimer’s and other neurological based disorders, including neuron response to medicine for faster drug testing.
My guess, 10,000x the cost on CL1. Even with the tech perfected, bio neurons fire much much slower than logic gates and electricity in a circuit board. If you have an ASIC (custom built board that isn’t really using a CPU), the ASIC would be much much faster for deterministic calculations at high speed with an active cooling system.
Bio neurons are great at self-organizing. If you already know how they need to be organized (e.g. a hashing algorithm), and you need max-speed output there’s no real advantage.
It’s not wrong to say bio neurons are power efficient, its just that power efficiency depends on what the activity is.
I guess that’s why in my circle of interest people are mainly using it for stuff like playing doom and controlling drone flight rather than hard number crunching