Ask HN: Is AI a Pseudoscience?

AI is no more one thing than cheese is or cancer is.

Plenty of good stuff is happening, but the reckless greed-fueled current hype part is not good.

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.

> Is AI a Pseudoscience?

That's like asking "Are Merkle-trees fake?"

Even if it gets posted during a week with a lot of other submissions about imploding cryptocurrencies, no amount of charitable-theorizing on my part can stop the answer from being a clear "No."