Why AI is difficult to discuss rationally

I think one of the reasons AI is difficult to discuss rationally is because the ramifications will be so wide-spread and impossible to predict. Currently we are seeing a lot of low effort results attached to a LOT of hype. If I had one bitcoin for every time someone evangailises AI without any evidence, I would have a lot of bitcoin. Here is an example of low effort, shared by Ted Goia:

So this is effectively a scam. And all the brand new experts promoting AI seem to gravitate towards low-hanging fruit. Why don’t we hear more about how AI and ML is great for detecting cancer or analysing things that humans cannot? Ripping off someones book, or their music or their art does not even seem like progress by comparison.

But here is a music comparison I think is maybe more worthy:

When drum machines were invented some people claimed it would make drummers redundant.
Of course it didn’t, but like AI it expands what is possible. Neither a drum machine nor AI will ever replace Jack de Johnette or Sly Dunbar or many other human drummers, because those are unique human beings that exist beyond any algorithm or computation.

But I think a better comparison is the TB303. Roland invented it with one specific intent. It failed to impress, much as AI generated material does now imho. But years later someone was playing with that same useless TB303 in a way no one predicted and they accidentally invented an entirely new genre of music!

There will unintended consequences with AI. Some of those consequences will be fantastic, and some will be destructive and potentially catastrophic. The only conclusion is no one knows – neither the cynics nor the evangelists. The only thing we can be sure of is that there will be unintended consequences





One thought on “Why AI is difficult to discuss rationally

  1. Richard

    Solid TB-303 analogy. I’m enjoying our current state of “artifacts as aesthetics” in the output of Midjourney, for example.

Leave a Reply

Your email address will not be published. Required fields are marked *