Exploring the Artificial Stupidity in Artificial Intelligence

Yes, This is an actual construct within the more significant AI Concept, and No, this is not a satirical article. The term "Artificial Stupidity" is usually used in two different contexts - "AI that makes mistakes which look silly to a human" or "AI that makes silly mistakes to look like a human".

Those two statements can evoke very different emotions in humans - the former is enjoyable and the latter, not so much (PS I've seen enough Terminator movies to imagine the worst-case scenarios). But returning to reality from fiction, we're not likely yet to face an AI armageddon, at least not anytime soon.

This news article about a bot getting fired from a supermarket serves as an excellent example for bots making mistakes in human context but seems to be correct in a literal sense. To tell a supermarket customer that "the Cheese is in the refrigerator" for someone searching the aisle containing the cheese may be correct in a literal sense but not exactly helpful in common sense.

Credits for above video - Jason Pheng - Original Source

These AI mistakes can be considered in the "inconvenience" bucket since they could be annoying but don't cause permanent damage. But, AI engines rejecting applications due to gender or racial reasons is concerning. ML models trained with biased datasets would prove to be very efficient for very wrong reasons. And these issues seem to crop up frequently in recent months. The silver lining is that we can correct these without losing the attractive benefits of Computer Vision, Machine Learning, etc.

Tackling the bias in AI head-on!

Creating transparency, i.e., losing Black box models in favor of explainable models, using non-biased training data, understanding the real-world impact with human reasoning, can be the baby steps to this problem.

Understanding the process used to deliver an outcome is often the starting point to improve the outcome quality.

Making better use of AI tools

These biases can be avoided by not making judgments just based on AI models. The models should be used to augment human reasoning and decision-making. This practice ensures human accountability for the AI models that impact human lives.

At the end of the day, AI is here to help humans make decisions and not make decisions for humans.

Let me know your thoughts on Artificial Stupidity in the comments below.

Disclaimer: These articles are merely my personal views and should not be construed or interpreted in any other way or form.