Is AI really that smart? Will AI take over the world? Are we going to see an incarnation of Skynet soon?
Maybe later. But not right now, AI hallucinates a lot. AI depends a lot on the information fed to them. They call it their “training”. Thus, the info also depends on the biases, culture, interests and availability of the information to the “trainers”.
Two years ago, I asked AI something about Moro (Muslims in the Philippines) history. It gave me a few paragraphs, with the usual authoritative tone. But I noticed that something was wrong. The sultans did not match with the century they lived or reigned. I told AI about it and it gave me different answers.
But many of the answers were wrong. Finally, it admitted that it did not know much about the history of “that corner of the world”.
Once, I asked if it knows astrology. It said it did. I asked something about Trump. It gave a long description of Trump’s astrological features.
But then I noticed that it gave the wrong Moon sign of Trump. I told AI abou it. It thanked me and changed Trump’s characteristics.
But then, I noticed other wrong things about his natal chart. I just stopped asking and concluded that this AI did not know astrology.
Two months later, a friend messaged me that AI did not know how to make natal charts. I answered that it does, but it makes wrong natal charts!
There are many instances that AI gave me the wrong answers. I also noticed that even some news features about recent historical events more than 20 or so years ago- whether online or print – were full of errors that could only occur if the writer had just read an old report about it and merely imagined the reasons for what was happening back then. I could only conclude that a young reporter found an easy way to write features on old events.
The reporter would just feed the AI a report from an old event, say an Olympic event, and let the AI write a story about it. The AI would then write what it imagined what logically could have happened at that time.
The editor, who also is not familiar with the topic, would just approve it. But those who lived and knew what exactly happened then could immediately see the discrepancies.
I have seen many news features in mainstream media that must have been written by AI.
I could write volumes about AI hallucinations.
But AI could also help in digging deeper into historical facts even if it started with the usual narrative. If your arguments are strong enough, it could dig much deeper to help you gain better understanding and have correct facts even if it means the AI will agree to your arguments and not to its original arguments.
Just be aware that AI hallucinates. And it hallucinates very authoritatively.
