Thoughts on AI

TL’DNR: I see warning signs everywhere around the topic. We need more honest discussion of AI, evaluation of pros and cons. For most areas, we should not yet invest our resources into GPUs, model training and inference.\nThese opinions and views are my own.\nResources AI is consuming resources, mostly\nfor production of the hardware for model training and inference, mostly GPUs for model training: for most LLMs, this is using up so much energy that just Google, Meta, IBM etc. can afford this for model inference, so “providing the AI infused service”, this uses up energy again Compared to the resources consumed in these areas, our brains are very efficient. Why not researching the technology more, and roll it out when we are sure the resources are taken into good use? We already run short on resources, and are fighting the climate crisis.\n