You may need to narrow your question. I think it's hard to deny at this point that there have been many proven outputs from the field of AI/deep learning - lang

You may need to narrow your question. I think it's hard to deny at this point that there have been many proven outputs from the field of AI/deep learning - language translation, defect detection, OCR, weather forecasting, spam filtering, protein folding, tumor segmentation, etc. Can say that there are also plenty of exaggerated claims about AI, but there is at least something real and useful at its core. Progress in the field appears largely based around empirical experimentation; researchers come up with some new technique (e.g: attention), set off runs with and without their proposed modification, then publish results. Not always as scientifically rigorous as it could be (no pre-registration, evaluation set leakage, etc.), but wouldn't say it's a pseudoscience.