Some thoughts from 2016 about the diminishing returns of IT technologies about AI:

Bitcoin Mining, self-propelled cars, as well as Quant Trading, are processing inputs to achieve an optimal result. As shown in the graphic, the decision is made in the 'Hidden' layer.

For neural networks (AI-BI) it means to go through 'lists' again and again and use the best results. This is how the machine - the automatic trader - learns the best results. But exactly this design principle
needs exponential more and more computing power and memory thus more and more (equal) current. Furthermore, this 'learning' process no longer ceases. It will become self-sufficient, the vending machines will become empty spaces and empty cars will be used for self-help:
'learning' of cars and maintenance robots or promotional, PR-AIs will watch us to 'learn'. Due to the competition in the relevant 'market', the income and benefits are becoming ever lower and the effort is ever higher.
It is as if more and more fishermen are fishing with better nets - for linear thinkers as we are - Paradoxical answer comes from fishing: less fishing increases yields. #artificialintelligent #ChatGPT https://www.pv4.eu/can-a-information-society-decouple-from-co2_1252.html
Can a Information Society decouple from CO2 – Photovoltaic for Europe