21.5 C
Pakistan
Saturday, June 15, 2024

AI may be a form of modern alchemy. That’s actually a good thing.

Let’s be modest about what we overlook while simultaneously being aware of what we could know.

The claim that AI is not scientific is one of the arguments that could harm it the most.

This is complicated since it defines ChatGPT and Midjourney in a pejorative manner, as if tacitly placing them at the bottom of the hierarchy of things that important to humanity, even while it doesn’t inherently discount the importance of what they can accomplish. Yes, they are useful, but what do they reveal about the outside world or ourselves? Nothing.

I have suggested that AI is best portrayed as an emerging science, which is not the same as being unscientific because it stresses the end result rather than the situation as it is right now. It’s also not nearly as awful. Even AI experts would not go as far as to call AI a hard science like physics or biology, but every field we revere now with almost religious fervor first emerged as a protoscience with aspirations comparable to the dubious techniques available at the time.

The term “aspiring science” is flattering. AI’s credibility comes under threat when it is implied that it doesn’t want to be a scientific enterprise by saying it isn’t one. If that were the case, alchemy and AI would be equivalent. That would be a significant issue.

Is AI the alchemy of our times?

The word “alchemy” carries a lot of baggage because of those who attempted—albeit unsuccessfully—to decisively distinguish it from chemistry. We instinctively believe that equating any subject of study with alchemy equates to labeling it as unserious — a topic destined to join astrology, humorism, and the aether ideas in the category of pseudoscience. However, I don’t believe the analogy between AI and alchemy should be taken in such a superficial, harsh way; if I’m being kind, I can see it as an attempt to call out the questionable research practices rather than to dismiss the legitimate prospects for it to become a science.

In that regard, I can understand the analogy: It is undeniable that contemporary deep learning was developed without a sound theoretical foundation, that milestones are attained through trial and error, and that the preferred way to advance is to feed data and computing power into the algorithms while working…

Related Articles

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Latest Articles