[EN] AI Winter II?

If another AI winter comes soon, it will be colder than the previous one.

Naturally, predicting an AI winter is non-trivial, but some signs seem to be present.

While today's LLMs have not exhausted their usefulness, their capabilities are limited and no further significant advance is visible yet (maybe one exists as a very well kept secret, but that is hard to believe, because such secrets are big and their numerous parts tend to become known).

Previous such advances have taken many years to materialize.

Besides, it is not a good sign that current AI technology requires so much brute force.

For example, the power consumption of a typical data center running AI is on the order of 100 MW (100,000,000 watts, while the power consumption of the human mind is on the order of 20 watts).

However, there are great expectations now, both by politicians and the general public.

And there is an abundance of hype.