In the defense of AI - we are not that flawless

The last couple of weeks have been spent talking about the incredible feats that AI have accomplished as well as how flawed it is. After the first wave of hype there has been a backlash where tech journalist after tech journalist points out the simple fact that AI is not flawless. In fact it’s flawed and often makes mistakes.

Google took a hit when their newly launched Bard AI claimed that the James Webb Space Telescope took the first picture of an exoplanet, a planet outside our own solar system, when it was asked to explain the JWST on a level a nine year old could understand.

This was, of course, incredibly wrong as the first picture was taken almost twenty years prior by another telescope. The AI was wrong, it had lied to us. The stock price tumbled and fell by seven percent in a day and the media caught on to the story. Soon there were numerous articles about how an AI could topple an empire as large as Google and how such an advanced model could be so wrong. How could something that expensive be so flawed?

Journalists and like minded techies all over the word started probing and taking jabs at all the language models they could get their hands on. More errors. More factual mistakes and strange answers emerged. The tidal wave of AI positive journalism had into a tsunami of AI skeptics “propaganda” overnight.

Before I continue I should probably take a stand on this question myself. To place myself into one of the two, binary, categories in this debate. The skeptics and the fanatics.

Having to choose I would say that I’m somewhere in between. I would like to call myself pragmatically AI positive. I think that there is much we still have to do and figure out until our high set standards and pipedreams about AI become reality, but we are getting there one step at a time. I’m also a skeptic, but not only of AI but of humanity itself. Before you run screaming I should point out that I’m not anti human. Humans are great!

But, we tend to forget that, like the saying goes:

To Err is Human; to forgive divine.

We are not flawless. In fact that is one of the things that makes us humans so great. We make mistakes all the time. But we learn and adapt, try to figure out what went wrong and try again. That is our superpower. That is what makes us what we are. I love that we are flawed.

But when it comes to our creations, the machinery we build, the system we construct and the AIs that we train on the knowledge and culture that is us we will not accept anything else but absolute perfection.

Perfection should arise out of the creations of flawed humans? It does not make any sense to me.

The quote above comes from the Enlightenment poet Alexander Pope but we seldom take the rest of the line into account. We as the creators should take responsibility, to factcheck and to forgive.

The internet is filled with countless articles written by us, the flawed humans. Some by journalists, scientists, great minds of this and prior generations. But also by complete idiots, crazy people and by armies of trolls. All have one thing in common. They’re human and they’re flawed. We all make mistakes and when we do we either doubledown on that mistake to cover up the tracks, being the stubborn apes that we are, or we take responsibility and try to make amends.

What was it really that tanked the stock and left a sour aftertaste in the mouth for Google and the whole industri? Was it the fact that their new cool AI made a mistake or the fact that no one seems to have fact checked the results before publishing it. The error was on the machine’s side, it was the flawed nature of man.