Yes, artificial intelligence (AI) is often wrong. Admittedly, this digital wunderkind can spout smooth sentences with amazing ease, but its results cannot be trusted. If you’re looking for facts, say by using an AI chatbot such as ChatGPT, watch out! (A chatbot is a computer program that uses AI to answer questions in verbal form, thus mimicking a human response.) If you would like to learn more about using ChatGPT, basic information can be found by clicking here.
The frequent misstatements of facts by AI have been publicized for some time, but thanks to my son, Greg, I now have a personal example of how distorted a response from ChatGPT can be. As an experiment, Greg asked ChatGPT to provide a positive review of my most recent post at writerken.com, the one I had titled What are the Dangers of Artificial Intelligence?. ChatGPT fired back in quick order and gave a glowing account of the post.
Facile prose
In facile prose, ChatGPT claimed to have read my post. It offered soothing and complimentary agreement on what I had written. But the entire job was phony baloney. It praised me for arguments I hadn’t made. In a curious way, the entire result reminded me of words coming out of the mouth of a Washington politician, that being a blast of self-serving hot air. Check it out. Here’s the chatbot’s entire response to Greg’s request:
If you haven’t read my post on the dangers of AI, you can find it by clicking here. My major point was that AI conceivably could lead to the extinction of human life, an eye-popping worry. After all, over 300 eminent experts had raised this alarm, as had countless others. But this monster point was unmentioned by ChatGPT. A colossal failure! (ChatGPT had been fed countless gigabytes of data, but obviously not my specific blot post). On the other hand, the chatbot patted me on the shoulder for raising issues I had not mentioned, those of job displacement and the loss of human empathy.
This is not to say that chatbots cannot be amazingly creative. They are accomplished writers of poetry and even novels, forms in which reality need not exist, and usually doesn’t. Recent reports indicate that Amazon’s Kindle store is being flooded with books written by ChatGPT. This brings up another looming danger of AI, and specifically of chatbots like ChatGPT (and one that ChatGPT credited me for saying, so I’ll get to that now). Chatbots might, and very likely will, put countless writers out of work, especially those contracted as writers for corporations to compose letters, advertising content, and miscellaneous company publications. A sad result.
Take home message
My take home message today is clear. Artificial intelligence is often wrong, sometimes spectacularly so. Somehow that comforts me.
Final observations:
I’ve noticed that readers rarely click on my links, such as when I refer to another website, or when I link back to earlier posts I’ve written on this site. I guarantee that these are not teasers designed to get you into trouble, nor are they in any way profitable to me. Rather they simply are trusted sources that I have found to be valuable as additional sources of information.
Stay tuned. I may go on a political rampage next. I’ve even got two obvious targets in mind.
Ken -keep writing about this -most of us do not really understand the details, or why there are errors. Ralph
I wish I had a real grasp of it, Ralph. I’m just circling around the periphery, amazed what computers can do, and simultaneously in utter awe that the digital world encompasses a form of “expertise” exceeding that of all human experts in every field of endeavor. When I imagine Orwell’s Big Brother as an all-knowing computer in the future, I feel shivers sliding along my spinal column. Could it happen? Maybe.