2020 BRN Discussion, page-13117

Dipped my toe into the water today...

IMO the question is if there is an application of this to an airgapped IoT environment that requires a feedback loop...

Typically when you train a model to perform a certain task, you cut off the learning layer and have it run standalone in inference mode without the need for further input/training data or a cloud/data center connection anyway. You may need to update it once a year possibly depending in the complexity of the task at hand...this would be the argument against

So... is there a problem dynamic enough that would require live training of the model and its inference mechanism in an airgapped environment (ie that would need to be done on the chip)? I can think of a few..

Coming from an IT/AI background myself I can see the big players (Honeywell for example) potentially needing this in their millions of IoT devices and sensors IF their software/requirements catch-up..this is the reason I personally see this as a long punt and will be selling down my other positions to try and accumulate BRN.

They are using the right frameworks (Tensorflow) and languages (Python) which will make things easy for those in the data science fields to play with their kit... what I hope to see is this new software center in India ramp up their API maturity.... If they can get just one of these major IoT players to 'need' their differentiators, then yeh, we are talking hundreds of millions of devices needing this, so major revenue..or they could just be snapped up by an nVidia for example..

All IMO DYOR..