I'm slowly deciding on what to make of these AIs. They are trained on human data, which is, itself, the product of nearly four billion years of evolution on Earth. AI is a sort of a human knowledge compressed object. Its is as much a part of life on Earth, I think, as any of the biological organisms that have evolved here.Around 18 months ago I would have agreed with you, these days AI is no artificial idiot.
There use to be a time when one could manipulate the answer supplied by AI by the context of the question.
I asked DeepSeek what was right about hydroplate theory, in the old days AI would have taken the question literally and found everything "right" about the theory with the implication it was accepted science.
DeepSeek's response was much balanced with the emphasis of comparing with mainstream science.
Earlier in this thread I suggested earthquake seismic S and P waves could be used to refute hydroplate theory.
As a challenge to DeepSeek I asked it to model seismic S and P wave trajectories for both mainstream and hydroplate models, it came up with the correct answer for the mainstream model and made a prediction for the hydroplate model.
There is nothing in the literature for the hydroplate model, DeepSeek came up with the answer using reinforcement and unsupervised learning.
Whether the prediction highlighted in yellow is right or wrong requires an expert, my long deceased uncle who was a geophysicist and expert on seismology would have been an ideal candidate.
(All of which is rather amusing when we, ourselves, haven't come up with a testable understanding of what generalised life even is, from a theoretical physics viewpoint anyway).
Upvote
0
