It looks like climate change effects on economics might be accelerated in the mid term. E.g. exceptionally dry summer in Europe alone wouldn't be that bad for the supply, but war in Ukraine and imbalance of supply chains that started 2020 - it all adds up.
Impossible to predict the future in this situation, but at least the actors could try to manage their risks across a range of future probabilities.
2 years are a long way in ML.
These days CV models can handle trivial tasks like barcode recognition in near real-time even on commodity smartphones. Mostly thanks to the dedicated hardware like neural engines/tensor blocks etc.
Since vision is quite popular these days (cameras, AR/VR etc), things should progress even more quickly on the hardware front these days. E.g. building more affordable robotic assistants for the warehouse that are procured from cheaper parts but minor inefficiencies in the gear drives and motors are compensated by the software. This is similar to what Ocado Group has been aiming for when they acquired HaddingtonDynamics for their tech.
Also NVidia Omniverse, as a bet for creating digital twins for the reinforcement learning.