2024 BrainChip Discussion, page-3598

Absolutely correct.

"Neuromorphic is a dimensional shift to the old ways of Von Neumann"

...and the problem is the learning curve.

Many learning curves:

Learn AI: While the concept of AI is not that hard to get your head around, applying it to a practical real world problem (regardless of the software or hardware you've chosen) and building something on the lab bench is a BIG topic.
Learn Akida: Akida comes along, and there is a whole new learning curve. What is it - and how do I use it? Just going from zero to building something that blinks an LED would take some doing. To then learn its underlying operating principals to get skilled enough to build something useful that can process an AI workload would be a 3 to 4 month learning curve for an experienced engineer. Then developing something robust ready to demonstrate to the people who sign the cheques - add another 12 months.
Evaluating alternatives: The people who write the cheques are going to ask "why Brainchip? what else have you looked at?"
Apply the learnings to something we can sell: OK - we are now skilled enough in AI and Akida to start thinking about developing a product. It's not until you understand the tools and have tried them out can you then start envisaging what you might apply them to.
Wait a minute - there's a better way to do this: Like learning any skill, you need to progress through BEGINNER -> NOVICE --> COMPETENT --> EXPERT --> MASTER. The way you use Akida to initially solve an AI workload problem won't be the same as you progress through the competency range. You will likely scrap the first approach and start again. Once you have real insight and familiarity you will see many, many better methods and potential applications. How many Brainchip customers thought they were ready to start at manufacturing an Akida based product, but the people in the lab said "STOP! we've worked out how to do this a much better way."

A very steep learning curve.